sample1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "1caff954-124f-4ae7-9fcf-161cede13e8d",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-27 12:35:20.575154: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import keras\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.preprocessing import image\n",
    "from keras.applications.mobilenet import preprocess_input\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "from model import mobilenet_v3_large_CA,mobilenet_v3_small_CA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "38cb1830-a8e1-429a-9d58-39b664d12cac",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Copying files: 54305 files [00:10, 5089.45 files/s]\n"
     ]
    }
   ],
   "source": [
    "import splitfolders\n",
    "splitfolders.ratio('plantvillage dataset/color', output=\"output\", seed=3407, ratio=(0.7, 0.1, 0.2), group_prefix=None) # default values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "bfbd8fe2-2340-4a53-ab31-bf91d83a3d1f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 48443 images belonging to 38 classes.\n"
     ]
    }
   ],
   "source": [
    "train_datagen=ImageDataGenerator(preprocessing_function=preprocess_input) #included in our dependencies\n",
    "\n",
    "train_generator=train_datagen.flow_from_directory('./output/train', # this is where you specify the path to the main data folder  \n",
    "                                                 target_size=(224,224),\n",
    "                                                  color_mode='rgb',\n",
    "                                                 batch_size=32,\n",
    "                                                 class_mode='categorical',\n",
    "                                                 shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "ce341ad1-4d8d-42f7-9261-1fe65186c0b8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 9864 images belonging to 38 classes.\n"
     ]
    }
   ],
   "source": [
    "val_datagen=ImageDataGenerator(preprocessing_function=preprocess_input) #included in our dependencies\n",
    "\n",
    "val_generator=train_datagen.flow_from_directory('./output/val', # this is where you specify the path to the main data folder\n",
    "                                                 target_size=(224,224),\n",
    "                                                 color_mode='rgb',\n",
    "                                                 batch_size=32,\n",
    "                                                 class_mode='categorical',\n",
    "                                                 shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "507169b9-d0a2-45c6-8eaa-a7da2f942f0d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 18953 images belonging to 38 classes.\n"
     ]
    }
   ],
   "source": [
    "test_datagen=ImageDataGenerator(preprocessing_function=preprocess_input) #included in our dependencies\n",
    "\n",
    "test_generator=train_datagen.flow_from_directory('./output/test', # this is where you specify the path to the main data folder\n",
    "                                                 target_size=(224,224),\n",
    "                                                 color_mode='rgb',\n",
    "                                                 batch_size=32,\n",
    "                                                 class_mode='categorical',\n",
    "                                                 shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c771da76-4d41-420b-bf8e-ee5a8e31c4c2",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-27 12:35:36.382461: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcuda.so.1\n",
      "2022-12-27 12:35:36.413110: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: \n",
      "pciBusID: 0000:3f:00.0 name: NVIDIA GeForce RTX 3090 computeCapability: 8.6\n",
      "coreClock: 1.695GHz coreCount: 82 deviceMemorySize: 23.70GiB deviceMemoryBandwidth: 871.81GiB/s\n",
      "2022-12-27 12:35:36.413150: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0\n",
      "2022-12-27 12:35:36.419904: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11\n",
      "2022-12-27 12:35:36.419994: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11\n",
      "2022-12-27 12:35:36.421158: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcufft.so.10\n",
      "2022-12-27 12:35:36.421939: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcurand.so.10\n",
      "2022-12-27 12:35:36.422989: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusolver.so.11\n",
      "2022-12-27 12:35:36.423946: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcusparse.so.11\n",
      "2022-12-27 12:35:36.424120: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8\n",
      "2022-12-27 12:35:36.424879: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0\n",
      "2022-12-27 12:35:36.425336: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 AVX512F FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2022-12-27 12:35:36.429791: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1733] Found device 0 with properties: \n",
      "pciBusID: 0000:3f:00.0 name: NVIDIA GeForce RTX 3090 computeCapability: 8.6\n",
      "coreClock: 1.695GHz coreCount: 82 deviceMemorySize: 23.70GiB deviceMemoryBandwidth: 871.81GiB/s\n",
      "2022-12-27 12:35:36.430408: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1871] Adding visible gpu devices: 0\n",
      "2022-12-27 12:35:36.430450: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudart.so.11.0\n",
      "2022-12-27 12:35:37.189051: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1258] Device interconnect StreamExecutor with strength 1 edge matrix:\n",
      "2022-12-27 12:35:37.189085: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1264]      0 \n",
      "2022-12-27 12:35:37.189093: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1277] 0:   N \n",
      "2022-12-27 12:35:37.191437: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1418] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 22318 MB memory) -> physical GPU (device: 0, name: NVIDIA GeForce RTX 3090, pci bus id: 0000:3f:00.0, compute capability: 8.6)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 96, 48) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(48,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization/gamma:0' shape=(48,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization/beta:0' shape=(48,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_1), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 48, 96) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_1), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(96,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_2), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 48, 96) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_2), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(96,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_3), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 240, 120) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_3), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(120,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_1), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_1/gamma:0' shape=(120,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_1/beta:0' shape=(120,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_4), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 120, 240) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_4), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(240,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_5), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 120, 240) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_5), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(240,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_6), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 240, 120) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_6), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(120,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_2), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_2/gamma:0' shape=(120,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_2/beta:0' shape=(120,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_7), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 120, 240) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_7), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(240,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_8), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 120, 240) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_8), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(240,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_9), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 120, 60) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_9), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(60,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_3), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_3/gamma:0' shape=(60,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_3/beta:0' shape=(60,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_10), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 60, 120) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_10), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(120,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_11), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 60, 120) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_11), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(120,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_12), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 144, 72) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_12), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(72,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_4), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_4/gamma:0' shape=(72,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_4/beta:0' shape=(72,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_13), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 72, 144) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_13), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(144,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_14), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 72, 144) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_14), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(144,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_15), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 288, 144) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_15), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(144,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_5), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_5/gamma:0' shape=(144,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_5/beta:0' shape=(144,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_16), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 144, 288) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_16), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(288,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_17), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 144, 288) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_17), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(288,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_18), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 576, 288) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_18), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(288,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_6), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_6/gamma:0' shape=(288,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_6/beta:0' shape=(288,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_19), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 288, 576) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_19), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(576,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_20), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 288, 576) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_20), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(576,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_21), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/kernel:0' shape=(1, 1, 576, 288) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_21), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name1/bias:0' shape=(288,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.compat.v1.nn.fused_batch_norm_7), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'batch_normalization_7/gamma:0' shape=(288,) dtype=float32>\n",
      "  <tf.Variable 'batch_normalization_7/beta:0' shape=(288,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_22), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/kernel:0' shape=(1, 1, 288, 576) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_22), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name2/bias:0' shape=(576,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.convolution_23), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/kernel:0' shape=(1, 1, 288, 576) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n",
      "WARNING:tensorflow:\n",
      "The following Variables were used a Lambda layer's call (tf.nn.bias_add_23), but\n",
      "are not present in its tracked objects:\n",
      "  <tf.Variable 'name3/bias:0' shape=(576,) dtype=float32>\n",
      "It is possible that this is intended behavior, but it is more likely\n",
      "an omission. This is a strong indication that this layer should be\n",
      "formulated as a subclassed Layer rather than a Lambda layer.\n"
     ]
    }
   ],
   "source": [
    "model = mobilenet_v3_small_CA(num_classes=38)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3f29b372-235d-4c4a-847e-e6db5457bffa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"MobilenetV3smallCA\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            [(None, 224, 224, 3) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "Conv (Conv2D)                   (None, 112, 112, 16) 432         input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "Conv/BatchNorm (BatchNormalizat (None, 112, 112, 16) 64          Conv[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "Conv/HardSwish (HardSwish)      (None, 112, 112, 16) 0           Conv/BatchNorm[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/depthwise/pad (Ze (None, 113, 113, 16) 0           Conv/HardSwish[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/depthwise (Depthw (None, 56, 56, 16)   144         expanded_conv/depthwise/pad[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/depthwise/BatchNo (None, 56, 56, 16)   64          expanded_conv/depthwise[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/depthwise/ReLU (R (None, 56, 56, 16)   0           expanded_conv/depthwise/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/squeeze_excite/Av (None, 16)           0           expanded_conv/depthwise/ReLU[0][0\n",
      "__________________________________________________________________________________________________\n",
      "reshape (Reshape)               (None, 1, 1, 16)     0           expanded_conv/squeeze_excite/AvgP\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/squeeze_excite/Co (None, 1, 1, 8)      136         reshape[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/squeeze_excite/Re (None, 1, 1, 8)      0           expanded_conv/squeeze_excite/Conv\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/squeeze_excite/Co (None, 1, 1, 16)     144         expanded_conv/squeeze_excite/Relu\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/squeeze_excite/Ha (None, 1, 1, 16)     0           expanded_conv/squeeze_excite/Conv\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/squeeze_excite/Mu (None, 56, 56, 16)   0           expanded_conv/depthwise/ReLU[0][0\n",
      "                                                                 expanded_conv/squeeze_excite/Hard\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/project (Conv2D)  (None, 56, 56, 16)   256         expanded_conv/squeeze_excite/Mul[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv/project/BatchNorm (None, 56, 56, 16)   64          expanded_conv/project[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/expand (Conv2D) (None, 56, 56, 72)   1152        expanded_conv/project/BatchNorm[0\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/expand/BatchNor (None, 56, 56, 72)   288         expanded_conv_1/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/expand/ReLU (Re (None, 56, 56, 72)   0           expanded_conv_1/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/depthwise/pad ( (None, 57, 57, 72)   0           expanded_conv_1/expand/ReLU[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/depthwise (Dept (None, 28, 28, 72)   648         expanded_conv_1/depthwise/pad[0][\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/depthwise/Batch (None, 28, 28, 72)   288         expanded_conv_1/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/depthwise/ReLU  (None, 28, 28, 72)   0           expanded_conv_1/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/project (Conv2D (None, 28, 28, 24)   1728        expanded_conv_1/depthwise/ReLU[0]\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_1/project/BatchNo (None, 28, 28, 24)   96          expanded_conv_1/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/expand (Conv2D) (None, 28, 28, 88)   2112        expanded_conv_1/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/expand/BatchNor (None, 28, 28, 88)   352         expanded_conv_2/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/expand/ReLU (Re (None, 28, 28, 88)   0           expanded_conv_2/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/depthwise (Dept (None, 28, 28, 88)   792         expanded_conv_2/expand/ReLU[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/depthwise/Batch (None, 28, 28, 88)   352         expanded_conv_2/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/depthwise/ReLU  (None, 28, 28, 88)   0           expanded_conv_2/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/project (Conv2D (None, 28, 28, 24)   2112        expanded_conv_2/depthwise/ReLU[0]\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/project/BatchNo (None, 28, 28, 24)   96          expanded_conv_2/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_2/Add (Add)       (None, 28, 28, 24)   0           expanded_conv_1/project/BatchNorm\n",
      "                                                                 expanded_conv_2/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/expand (Conv2D) (None, 28, 28, 96)   2304        expanded_conv_2/Add[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/expand/BatchNor (None, 28, 28, 96)   384         expanded_conv_3/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/expand/HardSwis (None, 28, 28, 96)   0           expanded_conv_3/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/depthwise/pad ( (None, 31, 31, 96)   0           expanded_conv_3/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/depthwise (Dept (None, 14, 14, 96)   2400        expanded_conv_3/depthwise/pad[0][\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/depthwise/Batch (None, 14, 14, 96)   384         expanded_conv_3/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/depthwise/HardS (None, 14, 14, 96)   0           expanded_conv_3/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean (TFOpLambda (None, 14, 96)       0           expanded_conv_3/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_1 (TFOpLamb (None, 14, 96)       0           expanded_conv_3/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat (TFOpLambda)          (None, 28, 96)       0           tf.math.reduce_mean[0][0]        \n",
      "                                                                 tf.math.reduce_mean_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape (TFOpLambda) (3,)                 0           tf.concat[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem (Slici ()                   0           tf.compat.v1.shape[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape (TFOpLambda)         (None, 1, 28, 96)    0           tf.concat[0][0]                  \n",
      "                                                                 tf.__operators__.getitem[0][0]   \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution (TFOpLambda)  (None, 1, 28, 48)    0           tf.reshape[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add (TFOpLambda)     (None, 1, 28, 48)    0           tf.nn.convolution[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 28, 48),  0           tf.nn.bias_add[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu (TFOpLambda)         (None, 1, 28, 48)    0           tf.compat.v1.nn.fused_batch_norm[\n",
      "__________________________________________________________________________________________________\n",
      "tf.split (TFOpLambda)           [(None, 1, 14, 48),  0           tf.nn.relu[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_1 (TFOpLambd (4,)                 0           tf.split[0][1]                   \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_1 (Sli ()                   0           tf.compat.v1.shape_1[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_1 (TFOpLambda (None, 1, 14, 96)    0           tf.split[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_1 (TFOpLambda)       (None, 14, 1, 48)    0           tf.split[0][1]                   \n",
      "                                                                 tf.__operators__.getitem_1[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_1 (TFOpLambda)   (None, 1, 14, 96)    0           tf.nn.convolution_1[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_2 (TFOpLambda (None, 14, 1, 96)    0           tf.reshape_1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid (TFOpLambda)    (None, 1, 14, 96)    0           tf.nn.bias_add_1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_2 (TFOpLambda)   (None, 14, 1, 96)    0           tf.nn.convolution_2[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply (TFOpLambda)   (None, 14, 14, 96)   0           expanded_conv_3/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_1 (TFOpLambda)  (None, 14, 1, 96)    0           tf.nn.bias_add_2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_1 (TFOpLambda) (None, 14, 14, 96)   0           tf.math.multiply[0][0]           \n",
      "                                                                 tf.math.sigmoid_1[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add (TFOpLambd (None, 14, 14, 96)   0           expanded_conv_3/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/project (Conv2D (None, 14, 14, 40)   3840        tf.__operators__.add[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_3/project/BatchNo (None, 14, 14, 40)   160         expanded_conv_3/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/expand (Conv2D) (None, 14, 14, 240)  9600        expanded_conv_3/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/expand/BatchNor (None, 14, 14, 240)  960         expanded_conv_4/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/expand/HardSwis (None, 14, 14, 240)  0           expanded_conv_4/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/depthwise (Dept (None, 14, 14, 240)  6000        expanded_conv_4/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/depthwise/Batch (None, 14, 14, 240)  960         expanded_conv_4/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/depthwise/HardS (None, 14, 14, 240)  0           expanded_conv_4/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_2 (TFOpLamb (None, 14, 240)      0           expanded_conv_4/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_3 (TFOpLamb (None, 14, 240)      0           expanded_conv_4/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_1 (TFOpLambda)        (None, 28, 240)      0           tf.math.reduce_mean_2[0][0]      \n",
      "                                                                 tf.math.reduce_mean_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_2 (TFOpLambd (3,)                 0           tf.concat_1[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_2 (Sli ()                   0           tf.compat.v1.shape_2[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_2 (TFOpLambda)       (None, 1, 28, 240)   0           tf.concat_1[0][0]                \n",
      "                                                                 tf.__operators__.getitem_2[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_3 (TFOpLambda (None, 1, 28, 120)   0           tf.reshape_2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_3 (TFOpLambda)   (None, 1, 28, 120)   0           tf.nn.convolution_3[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 28, 120), 0           tf.nn.bias_add_3[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_1 (TFOpLambda)       (None, 1, 28, 120)   0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_1 (TFOpLambda)         [(None, 1, 14, 120), 0           tf.nn.relu_1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_3 (TFOpLambd (4,)                 0           tf.split_1[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_3 (Sli ()                   0           tf.compat.v1.shape_3[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_4 (TFOpLambda (None, 1, 14, 240)   0           tf.split_1[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_3 (TFOpLambda)       (None, 14, 1, 120)   0           tf.split_1[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_3[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_4 (TFOpLambda)   (None, 1, 14, 240)   0           tf.nn.convolution_4[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_5 (TFOpLambda (None, 14, 1, 240)   0           tf.reshape_3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_2 (TFOpLambda)  (None, 1, 14, 240)   0           tf.nn.bias_add_4[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_5 (TFOpLambda)   (None, 14, 1, 240)   0           tf.nn.convolution_5[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_2 (TFOpLambda) (None, 14, 14, 240)  0           expanded_conv_4/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid_2[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_3 (TFOpLambda)  (None, 14, 1, 240)   0           tf.nn.bias_add_5[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_3 (TFOpLambda) (None, 14, 14, 240)  0           tf.math.multiply_2[0][0]         \n",
      "                                                                 tf.math.sigmoid_3[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_1 (TFOpLam (None, 14, 14, 240)  0           expanded_conv_4/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_3[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/project (Conv2D (None, 14, 14, 40)   9600        tf.__operators__.add_1[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/project/BatchNo (None, 14, 14, 40)   160         expanded_conv_4/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_4/Add (Add)       (None, 14, 14, 40)   0           expanded_conv_3/project/BatchNorm\n",
      "                                                                 expanded_conv_4/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/expand (Conv2D) (None, 14, 14, 240)  9600        expanded_conv_4/Add[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/expand/BatchNor (None, 14, 14, 240)  960         expanded_conv_5/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/expand/HardSwis (None, 14, 14, 240)  0           expanded_conv_5/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/depthwise (Dept (None, 14, 14, 240)  6000        expanded_conv_5/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/depthwise/Batch (None, 14, 14, 240)  960         expanded_conv_5/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/depthwise/HardS (None, 14, 14, 240)  0           expanded_conv_5/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_4 (TFOpLamb (None, 14, 240)      0           expanded_conv_5/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_5 (TFOpLamb (None, 14, 240)      0           expanded_conv_5/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_2 (TFOpLambda)        (None, 28, 240)      0           tf.math.reduce_mean_4[0][0]      \n",
      "                                                                 tf.math.reduce_mean_5[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_4 (TFOpLambd (3,)                 0           tf.concat_2[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_4 (Sli ()                   0           tf.compat.v1.shape_4[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_4 (TFOpLambda)       (None, 1, 28, 240)   0           tf.concat_2[0][0]                \n",
      "                                                                 tf.__operators__.getitem_4[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_6 (TFOpLambda (None, 1, 28, 120)   0           tf.reshape_4[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_6 (TFOpLambda)   (None, 1, 28, 120)   0           tf.nn.convolution_6[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 28, 120), 0           tf.nn.bias_add_6[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_2 (TFOpLambda)       (None, 1, 28, 120)   0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_2 (TFOpLambda)         [(None, 1, 14, 120), 0           tf.nn.relu_2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_5 (TFOpLambd (4,)                 0           tf.split_2[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_5 (Sli ()                   0           tf.compat.v1.shape_5[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_7 (TFOpLambda (None, 1, 14, 240)   0           tf.split_2[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_5 (TFOpLambda)       (None, 14, 1, 120)   0           tf.split_2[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_5[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_7 (TFOpLambda)   (None, 1, 14, 240)   0           tf.nn.convolution_7[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_8 (TFOpLambda (None, 14, 1, 240)   0           tf.reshape_5[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_4 (TFOpLambda)  (None, 1, 14, 240)   0           tf.nn.bias_add_7[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_8 (TFOpLambda)   (None, 14, 1, 240)   0           tf.nn.convolution_8[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_4 (TFOpLambda) (None, 14, 14, 240)  0           expanded_conv_5/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid_4[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_5 (TFOpLambda)  (None, 14, 1, 240)   0           tf.nn.bias_add_8[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_5 (TFOpLambda) (None, 14, 14, 240)  0           tf.math.multiply_4[0][0]         \n",
      "                                                                 tf.math.sigmoid_5[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_2 (TFOpLam (None, 14, 14, 240)  0           expanded_conv_5/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_5[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/project (Conv2D (None, 14, 14, 40)   9600        tf.__operators__.add_2[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/project/BatchNo (None, 14, 14, 40)   160         expanded_conv_5/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_5/Add (Add)       (None, 14, 14, 40)   0           expanded_conv_4/Add[0][0]        \n",
      "                                                                 expanded_conv_5/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/expand (Conv2D) (None, 14, 14, 120)  4800        expanded_conv_5/Add[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/expand/BatchNor (None, 14, 14, 120)  480         expanded_conv_6/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/expand/HardSwis (None, 14, 14, 120)  0           expanded_conv_6/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/depthwise (Dept (None, 14, 14, 120)  3000        expanded_conv_6/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/depthwise/Batch (None, 14, 14, 120)  480         expanded_conv_6/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/depthwise/HardS (None, 14, 14, 120)  0           expanded_conv_6/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_6 (TFOpLamb (None, 14, 120)      0           expanded_conv_6/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_7 (TFOpLamb (None, 14, 120)      0           expanded_conv_6/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_3 (TFOpLambda)        (None, 28, 120)      0           tf.math.reduce_mean_6[0][0]      \n",
      "                                                                 tf.math.reduce_mean_7[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_6 (TFOpLambd (3,)                 0           tf.concat_3[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_6 (Sli ()                   0           tf.compat.v1.shape_6[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_6 (TFOpLambda)       (None, 1, 28, 120)   0           tf.concat_3[0][0]                \n",
      "                                                                 tf.__operators__.getitem_6[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_9 (TFOpLambda (None, 1, 28, 60)    0           tf.reshape_6[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_9 (TFOpLambda)   (None, 1, 28, 60)    0           tf.nn.convolution_9[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 28, 60),  0           tf.nn.bias_add_9[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_3 (TFOpLambda)       (None, 1, 28, 60)    0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_3 (TFOpLambda)         [(None, 1, 14, 60),  0           tf.nn.relu_3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_7 (TFOpLambd (4,)                 0           tf.split_3[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_7 (Sli ()                   0           tf.compat.v1.shape_7[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_10 (TFOpLambd (None, 1, 14, 120)   0           tf.split_3[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_7 (TFOpLambda)       (None, 14, 1, 60)    0           tf.split_3[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_7[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_10 (TFOpLambda)  (None, 1, 14, 120)   0           tf.nn.convolution_10[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_11 (TFOpLambd (None, 14, 1, 120)   0           tf.reshape_7[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_6 (TFOpLambda)  (None, 1, 14, 120)   0           tf.nn.bias_add_10[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_11 (TFOpLambda)  (None, 14, 1, 120)   0           tf.nn.convolution_11[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_6 (TFOpLambda) (None, 14, 14, 120)  0           expanded_conv_6/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid_6[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_7 (TFOpLambda)  (None, 14, 1, 120)   0           tf.nn.bias_add_11[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_7 (TFOpLambda) (None, 14, 14, 120)  0           tf.math.multiply_6[0][0]         \n",
      "                                                                 tf.math.sigmoid_7[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_3 (TFOpLam (None, 14, 14, 120)  0           expanded_conv_6/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_7[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/project (Conv2D (None, 14, 14, 48)   5760        tf.__operators__.add_3[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_6/project/BatchNo (None, 14, 14, 48)   192         expanded_conv_6/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/expand (Conv2D) (None, 14, 14, 144)  6912        expanded_conv_6/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/expand/BatchNor (None, 14, 14, 144)  576         expanded_conv_7/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/expand/HardSwis (None, 14, 14, 144)  0           expanded_conv_7/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/depthwise (Dept (None, 14, 14, 144)  3600        expanded_conv_7/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/depthwise/Batch (None, 14, 14, 144)  576         expanded_conv_7/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/depthwise/HardS (None, 14, 14, 144)  0           expanded_conv_7/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_8 (TFOpLamb (None, 14, 144)      0           expanded_conv_7/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_9 (TFOpLamb (None, 14, 144)      0           expanded_conv_7/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_4 (TFOpLambda)        (None, 28, 144)      0           tf.math.reduce_mean_8[0][0]      \n",
      "                                                                 tf.math.reduce_mean_9[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_8 (TFOpLambd (3,)                 0           tf.concat_4[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_8 (Sli ()                   0           tf.compat.v1.shape_8[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_8 (TFOpLambda)       (None, 1, 28, 144)   0           tf.concat_4[0][0]                \n",
      "                                                                 tf.__operators__.getitem_8[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_12 (TFOpLambd (None, 1, 28, 72)    0           tf.reshape_8[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_12 (TFOpLambda)  (None, 1, 28, 72)    0           tf.nn.convolution_12[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 28, 72),  0           tf.nn.bias_add_12[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_4 (TFOpLambda)       (None, 1, 28, 72)    0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_4 (TFOpLambda)         [(None, 1, 14, 72),  0           tf.nn.relu_4[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_9 (TFOpLambd (4,)                 0           tf.split_4[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_9 (Sli ()                   0           tf.compat.v1.shape_9[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_13 (TFOpLambd (None, 1, 14, 144)   0           tf.split_4[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_9 (TFOpLambda)       (None, 14, 1, 72)    0           tf.split_4[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_9[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_13 (TFOpLambda)  (None, 1, 14, 144)   0           tf.nn.convolution_13[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_14 (TFOpLambd (None, 14, 1, 144)   0           tf.reshape_9[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_8 (TFOpLambda)  (None, 1, 14, 144)   0           tf.nn.bias_add_13[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_14 (TFOpLambda)  (None, 14, 1, 144)   0           tf.nn.convolution_14[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_8 (TFOpLambda) (None, 14, 14, 144)  0           expanded_conv_7/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid_8[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_9 (TFOpLambda)  (None, 14, 1, 144)   0           tf.nn.bias_add_14[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_9 (TFOpLambda) (None, 14, 14, 144)  0           tf.math.multiply_8[0][0]         \n",
      "                                                                 tf.math.sigmoid_9[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_4 (TFOpLam (None, 14, 14, 144)  0           expanded_conv_7/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_9[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/project (Conv2D (None, 14, 14, 48)   6912        tf.__operators__.add_4[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/project/BatchNo (None, 14, 14, 48)   192         expanded_conv_7/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_7/Add (Add)       (None, 14, 14, 48)   0           expanded_conv_6/project/BatchNorm\n",
      "                                                                 expanded_conv_7/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/expand (Conv2D) (None, 14, 14, 288)  13824       expanded_conv_7/Add[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/expand/BatchNor (None, 14, 14, 288)  1152        expanded_conv_8/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/expand/HardSwis (None, 14, 14, 288)  0           expanded_conv_8/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/depthwise/pad ( (None, 17, 17, 288)  0           expanded_conv_8/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/depthwise (Dept (None, 7, 7, 288)    7200        expanded_conv_8/depthwise/pad[0][\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/depthwise/Batch (None, 7, 7, 288)    1152        expanded_conv_8/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/depthwise/HardS (None, 7, 7, 288)    0           expanded_conv_8/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_10 (TFOpLam (None, 7, 288)       0           expanded_conv_8/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_11 (TFOpLam (None, 7, 288)       0           expanded_conv_8/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_5 (TFOpLambda)        (None, 14, 288)      0           tf.math.reduce_mean_10[0][0]     \n",
      "                                                                 tf.math.reduce_mean_11[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_10 (TFOpLamb (3,)                 0           tf.concat_5[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_10 (Sl ()                   0           tf.compat.v1.shape_10[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_10 (TFOpLambda)      (None, 1, 14, 288)   0           tf.concat_5[0][0]                \n",
      "                                                                 tf.__operators__.getitem_10[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_15 (TFOpLambd (None, 1, 14, 144)   0           tf.reshape_10[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_15 (TFOpLambda)  (None, 1, 14, 144)   0           tf.nn.convolution_15[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 14, 144), 0           tf.nn.bias_add_15[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_5 (TFOpLambda)       (None, 1, 14, 144)   0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_5 (TFOpLambda)         [(None, 1, 7, 144),  0           tf.nn.relu_5[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_11 (TFOpLamb (4,)                 0           tf.split_5[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_11 (Sl ()                   0           tf.compat.v1.shape_11[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_16 (TFOpLambd (None, 1, 7, 288)    0           tf.split_5[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_11 (TFOpLambda)      (None, 7, 1, 144)    0           tf.split_5[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_11[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_16 (TFOpLambda)  (None, 1, 7, 288)    0           tf.nn.convolution_16[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_17 (TFOpLambd (None, 7, 1, 288)    0           tf.reshape_11[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_10 (TFOpLambda) (None, 1, 7, 288)    0           tf.nn.bias_add_16[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_17 (TFOpLambda)  (None, 7, 1, 288)    0           tf.nn.convolution_17[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_10 (TFOpLambda (None, 7, 7, 288)    0           expanded_conv_8/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid_10[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_11 (TFOpLambda) (None, 7, 1, 288)    0           tf.nn.bias_add_17[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_11 (TFOpLambda (None, 7, 7, 288)    0           tf.math.multiply_10[0][0]        \n",
      "                                                                 tf.math.sigmoid_11[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_5 (TFOpLam (None, 7, 7, 288)    0           expanded_conv_8/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_11[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/project (Conv2D (None, 7, 7, 96)     27648       tf.__operators__.add_5[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_8/project/BatchNo (None, 7, 7, 96)     384         expanded_conv_8/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/expand (Conv2D) (None, 7, 7, 576)    55296       expanded_conv_8/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/expand/BatchNor (None, 7, 7, 576)    2304        expanded_conv_9/expand[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/expand/HardSwis (None, 7, 7, 576)    0           expanded_conv_9/expand/BatchNorm[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/depthwise (Dept (None, 7, 7, 576)    14400       expanded_conv_9/expand/HardSwish[\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/depthwise/Batch (None, 7, 7, 576)    2304        expanded_conv_9/depthwise[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/depthwise/HardS (None, 7, 7, 576)    0           expanded_conv_9/depthwise/BatchNo\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_12 (TFOpLam (None, 7, 576)       0           expanded_conv_9/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_13 (TFOpLam (None, 7, 576)       0           expanded_conv_9/depthwise/HardSwi\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_6 (TFOpLambda)        (None, 14, 576)      0           tf.math.reduce_mean_12[0][0]     \n",
      "                                                                 tf.math.reduce_mean_13[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_12 (TFOpLamb (3,)                 0           tf.concat_6[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_12 (Sl ()                   0           tf.compat.v1.shape_12[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_12 (TFOpLambda)      (None, 1, 14, 576)   0           tf.concat_6[0][0]                \n",
      "                                                                 tf.__operators__.getitem_12[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_18 (TFOpLambd (None, 1, 14, 288)   0           tf.reshape_12[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_18 (TFOpLambda)  (None, 1, 14, 288)   0           tf.nn.convolution_18[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 14, 288), 0           tf.nn.bias_add_18[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_6 (TFOpLambda)       (None, 1, 14, 288)   0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_6 (TFOpLambda)         [(None, 1, 7, 288),  0           tf.nn.relu_6[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_13 (TFOpLamb (4,)                 0           tf.split_6[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_13 (Sl ()                   0           tf.compat.v1.shape_13[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_19 (TFOpLambd (None, 1, 7, 576)    0           tf.split_6[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_13 (TFOpLambda)      (None, 7, 1, 288)    0           tf.split_6[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_13[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_19 (TFOpLambda)  (None, 1, 7, 576)    0           tf.nn.convolution_19[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_20 (TFOpLambd (None, 7, 1, 576)    0           tf.reshape_13[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_12 (TFOpLambda) (None, 1, 7, 576)    0           tf.nn.bias_add_19[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_20 (TFOpLambda)  (None, 7, 1, 576)    0           tf.nn.convolution_20[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_12 (TFOpLambda (None, 7, 7, 576)    0           expanded_conv_9/depthwise/HardSwi\n",
      "                                                                 tf.math.sigmoid_12[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_13 (TFOpLambda) (None, 7, 1, 576)    0           tf.nn.bias_add_20[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_13 (TFOpLambda (None, 7, 7, 576)    0           tf.math.multiply_12[0][0]        \n",
      "                                                                 tf.math.sigmoid_13[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_6 (TFOpLam (None, 7, 7, 576)    0           expanded_conv_9/depthwise/HardSwi\n",
      "                                                                 tf.math.multiply_13[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/project (Conv2D (None, 7, 7, 96)     55296       tf.__operators__.add_6[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/project/BatchNo (None, 7, 7, 96)     384         expanded_conv_9/project[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_9/Add (Add)       (None, 7, 7, 96)     0           expanded_conv_8/project/BatchNorm\n",
      "                                                                 expanded_conv_9/project/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/expand (Conv2D (None, 7, 7, 576)    55296       expanded_conv_9/Add[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/expand/BatchNo (None, 7, 7, 576)    2304        expanded_conv_10/expand[0][0]    \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/expand/HardSwi (None, 7, 7, 576)    0           expanded_conv_10/expand/BatchNorm\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/depthwise (Dep (None, 7, 7, 576)    14400       expanded_conv_10/expand/HardSwish\n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/depthwise/Batc (None, 7, 7, 576)    2304        expanded_conv_10/depthwise[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/depthwise/Hard (None, 7, 7, 576)    0           expanded_conv_10/depthwise/BatchN\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_14 (TFOpLam (None, 7, 576)       0           expanded_conv_10/depthwise/HardSw\n",
      "__________________________________________________________________________________________________\n",
      "tf.math.reduce_mean_15 (TFOpLam (None, 7, 576)       0           expanded_conv_10/depthwise/HardSw\n",
      "__________________________________________________________________________________________________\n",
      "tf.concat_7 (TFOpLambda)        (None, 14, 576)      0           tf.math.reduce_mean_14[0][0]     \n",
      "                                                                 tf.math.reduce_mean_15[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_14 (TFOpLamb (3,)                 0           tf.concat_7[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_14 (Sl ()                   0           tf.compat.v1.shape_14[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_14 (TFOpLambda)      (None, 1, 14, 576)   0           tf.concat_7[0][0]                \n",
      "                                                                 tf.__operators__.getitem_14[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_21 (TFOpLambd (None, 1, 14, 288)   0           tf.reshape_14[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_21 (TFOpLambda)  (None, 1, 14, 288)   0           tf.nn.convolution_21[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.nn.fused_batch_nor ((None, 1, 14, 288), 0           tf.nn.bias_add_21[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.relu_7 (TFOpLambda)       (None, 1, 14, 288)   0           tf.compat.v1.nn.fused_batch_norm_\n",
      "__________________________________________________________________________________________________\n",
      "tf.split_7 (TFOpLambda)         [(None, 1, 7, 288),  0           tf.nn.relu_7[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "tf.compat.v1.shape_15 (TFOpLamb (4,)                 0           tf.split_7[0][1]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.getitem_15 (Sl ()                   0           tf.compat.v1.shape_15[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_22 (TFOpLambd (None, 1, 7, 576)    0           tf.split_7[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "tf.reshape_15 (TFOpLambda)      (None, 7, 1, 288)    0           tf.split_7[0][1]                 \n",
      "                                                                 tf.__operators__.getitem_15[0][0]\n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_22 (TFOpLambda)  (None, 1, 7, 576)    0           tf.nn.convolution_22[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.convolution_23 (TFOpLambd (None, 7, 1, 576)    0           tf.reshape_15[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_14 (TFOpLambda) (None, 1, 7, 576)    0           tf.nn.bias_add_22[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.nn.bias_add_23 (TFOpLambda)  (None, 7, 1, 576)    0           tf.nn.convolution_23[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_14 (TFOpLambda (None, 7, 7, 576)    0           expanded_conv_10/depthwise/HardSw\n",
      "                                                                 tf.math.sigmoid_14[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.sigmoid_15 (TFOpLambda) (None, 7, 1, 576)    0           tf.nn.bias_add_23[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "tf.math.multiply_15 (TFOpLambda (None, 7, 7, 576)    0           tf.math.multiply_14[0][0]        \n",
      "                                                                 tf.math.sigmoid_15[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "tf.__operators__.add_7 (TFOpLam (None, 7, 7, 576)    0           expanded_conv_10/depthwise/HardSw\n",
      "                                                                 tf.math.multiply_15[0][0]        \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/project (Conv2 (None, 7, 7, 96)     55296       tf.__operators__.add_7[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/project/BatchN (None, 7, 7, 96)     384         expanded_conv_10/project[0][0]   \n",
      "__________________________________________________________________________________________________\n",
      "expanded_conv_10/Add (Add)      (None, 7, 7, 96)     0           expanded_conv_9/Add[0][0]        \n",
      "                                                                 expanded_conv_10/project/BatchNor\n",
      "__________________________________________________________________________________________________\n",
      "Conv_1 (Conv2D)                 (None, 7, 7, 576)    55296       expanded_conv_10/Add[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "Conv_1/BatchNorm (BatchNormaliz (None, 7, 7, 576)    2304        Conv_1[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "Conv_1/HardSwish (HardSwish)    (None, 7, 7, 576)    0           Conv_1/BatchNorm[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d (Globa (None, 576)          0           Conv_1/HardSwish[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "reshape_1 (Reshape)             (None, 1, 1, 576)    0           global_average_pooling2d[0][0]   \n",
      "__________________________________________________________________________________________________\n",
      "Conv_2 (Conv2D)                 (None, 1, 1, 1024)   590848      reshape_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "Conv_2/HardSwish (HardSwish)    (None, 1, 1, 1024)   0           Conv_2[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "Logits/Conv2d_1c_1x1 (Conv2D)   (None, 1, 1, 38)     38950       Conv_2/HardSwish[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "flatten (Flatten)               (None, 38)           0           Logits/Conv2d_1c_1x1[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "Predictions (Softmax)           (None, 38)           0           flatten[0][0]                    \n",
      "==================================================================================================\n",
      "Total params: 1,107,558\n",
      "Trainable params: 1,095,446\n",
      "Non-trainable params: 12,112\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "6222f22e-c950-4675-a6a8-768246c8ff27",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/miniconda3/lib/python3.8/site-packages/tensorflow/python/keras/engine/training.py:1940: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
      "  warnings.warn('`Model.fit_generator` is deprecated and '\n",
      "2022-12-27 12:35:39.520533: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:176] None of the MLIR Optimization Passes are enabled (registered 2)\n",
      "2022-12-27 12:35:39.521140: I tensorflow/core/platform/profile_utils/cpu_utils.cc:114] CPU Frequency: 2500000000 Hz\n",
      "2022-12-27 12:35:46.317997: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcudnn.so.8\n",
      "2022-12-27 12:35:47.273075: I tensorflow/stream_executor/cuda/cuda_dnn.cc:359] Loaded cuDNN version 8101\n",
      "2022-12-27 12:35:48.412771: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublas.so.11\n",
      "2022-12-27 12:35:49.279085: I tensorflow/stream_executor/platform/default/dso_loader.cc:53] Successfully opened dynamic library libcublasLt.so.11\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1513/1513 [==============================] - 151s 93ms/step - loss: 1.0009 - accuracy: 0.7000 - val_loss: 0.6700 - val_accuracy: 0.8006\n",
      "Epoch 2/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.3398 - accuracy: 0.8880 - val_loss: 0.2787 - val_accuracy: 0.9068\n",
      "Epoch 3/30\n",
      "1513/1513 [==============================] - 137s 90ms/step - loss: 0.2257 - accuracy: 0.9262 - val_loss: 0.3164 - val_accuracy: 0.8971\n",
      "Epoch 4/30\n",
      "1513/1513 [==============================] - 137s 91ms/step - loss: 0.1813 - accuracy: 0.9411 - val_loss: 0.3561 - val_accuracy: 0.8944\n",
      "Epoch 5/30\n",
      "1513/1513 [==============================] - 138s 91ms/step - loss: 0.1430 - accuracy: 0.9536 - val_loss: 0.1278 - val_accuracy: 0.9543\n",
      "Epoch 6/30\n",
      "1513/1513 [==============================] - 137s 91ms/step - loss: 0.1217 - accuracy: 0.9594 - val_loss: 0.1983 - val_accuracy: 0.9393\n",
      "Epoch 7/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.1064 - accuracy: 0.9651 - val_loss: 0.1365 - val_accuracy: 0.9570\n",
      "Epoch 8/30\n",
      "1513/1513 [==============================] - 137s 90ms/step - loss: 0.0933 - accuracy: 0.9690 - val_loss: 0.2245 - val_accuracy: 0.9303\n",
      "Epoch 9/30\n",
      "1513/1513 [==============================] - 137s 90ms/step - loss: 0.0858 - accuracy: 0.9723 - val_loss: 0.1464 - val_accuracy: 0.9553\n",
      "Epoch 10/30\n",
      "1513/1513 [==============================] - 137s 90ms/step - loss: 0.0746 - accuracy: 0.9754 - val_loss: 0.4588 - val_accuracy: 0.8968\n",
      "Epoch 11/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0663 - accuracy: 0.9781 - val_loss: 0.1023 - val_accuracy: 0.9719\n",
      "Epoch 12/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0666 - accuracy: 0.9782 - val_loss: 0.0590 - val_accuracy: 0.9811\n",
      "Epoch 13/30\n",
      "1513/1513 [==============================] - 138s 91ms/step - loss: 0.0561 - accuracy: 0.9818 - val_loss: 0.1705 - val_accuracy: 0.9513\n",
      "Epoch 14/30\n",
      "1513/1513 [==============================] - 138s 91ms/step - loss: 0.0577 - accuracy: 0.9822 - val_loss: 0.0597 - val_accuracy: 0.9826\n",
      "Epoch 15/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0520 - accuracy: 0.9826 - val_loss: 0.0493 - val_accuracy: 0.9866\n",
      "Epoch 16/30\n",
      "1513/1513 [==============================] - 137s 91ms/step - loss: 0.0508 - accuracy: 0.9829 - val_loss: 0.1392 - val_accuracy: 0.9608\n",
      "Epoch 17/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0450 - accuracy: 0.9860 - val_loss: 0.1173 - val_accuracy: 0.9680\n",
      "Epoch 18/30\n",
      "1513/1513 [==============================] - 137s 90ms/step - loss: 0.0452 - accuracy: 0.9856 - val_loss: 0.0856 - val_accuracy: 0.9738\n",
      "Epoch 19/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0420 - accuracy: 0.9858 - val_loss: 0.0496 - val_accuracy: 0.9858\n",
      "Epoch 20/30\n",
      "1513/1513 [==============================] - 137s 90ms/step - loss: 0.0390 - accuracy: 0.9872 - val_loss: 0.0387 - val_accuracy: 0.9900\n",
      "Epoch 21/30\n",
      "1513/1513 [==============================] - 137s 91ms/step - loss: 0.0375 - accuracy: 0.9880 - val_loss: 0.0586 - val_accuracy: 0.9845\n",
      "Epoch 22/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0333 - accuracy: 0.9895 - val_loss: 0.0723 - val_accuracy: 0.9789\n",
      "Epoch 23/30\n",
      "1513/1513 [==============================] - 137s 91ms/step - loss: 0.0366 - accuracy: 0.9887 - val_loss: 0.2729 - val_accuracy: 0.9413\n",
      "Epoch 24/30\n",
      "1513/1513 [==============================] - 136s 90ms/step - loss: 0.0331 - accuracy: 0.9897 - val_loss: 0.0404 - val_accuracy: 0.9872\n",
      "Epoch 25/30\n",
      "1513/1513 [==============================] - 137s 91ms/step - loss: 0.0304 - accuracy: 0.9906 - val_loss: 0.0625 - val_accuracy: 0.9819\n",
      "Epoch 26/30\n",
      "1513/1513 [==============================] - 144s 95ms/step - loss: 0.0338 - accuracy: 0.9896 - val_loss: 0.0675 - val_accuracy: 0.9800\n",
      "Epoch 27/30\n",
      "1513/1513 [==============================] - 146s 96ms/step - loss: 0.0306 - accuracy: 0.9907 - val_loss: 0.0449 - val_accuracy: 0.9870\n",
      "Epoch 28/30\n",
      "1513/1513 [==============================] - 145s 96ms/step - loss: 0.0283 - accuracy: 0.9910 - val_loss: 0.0644 - val_accuracy: 0.9861\n",
      "Epoch 29/30\n",
      "1513/1513 [==============================] - 146s 96ms/step - loss: 0.0289 - accuracy: 0.9915 - val_loss: 0.0296 - val_accuracy: 0.9903\n",
      "Epoch 30/30\n",
      "1513/1513 [==============================] - 146s 97ms/step - loss: 0.0299 - accuracy: 0.9916 - val_loss: 0.0276 - val_accuracy: 0.9925\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='Adam',loss='categorical_crossentropy',metrics=['accuracy'])\n",
    "# Adam optimizer\n",
    "# loss function will be categorical cross entropy\n",
    "# evaluation metric will be accuracy\n",
    "\n",
    "step_size_train=train_generator.n//train_generator.batch_size\n",
    "accident_model=model.fit_generator(generator=train_generator,\n",
    "                    validation_data =val_generator,\n",
    "                   steps_per_epoch=step_size_train,\n",
    "                   epochs=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "f7f1be3f-9a9a-4b5d-b5d5-91c272d8cc4f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f4278387f70>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#train_visualization\n",
    "plt.plot(accident_model.history['loss'], label = 'training loss')\n",
    "plt.plot(accident_model.history['accuracy'], label = 'training accuracy')\n",
    "plt.grid(True)\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "de36a322-af71-4730-8d88-b7b47dfcc1ee",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f42782eb610>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#vaild_visualization\n",
    "plt.plot(accident_model.history['val_loss'], label = 'validation loss')\n",
    "plt.plot(accident_model.history['val_accuracy'], label = 'validation accuracy')\n",
    "plt.grid(True)\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "d55bee17-08bf-4244-8cac-6589f5aac6a2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "24/24 [==============================] - 2s 87ms/step - loss: 0.0182 - accuracy: 0.9961\n",
      "Test score: 0.01824832707643509\n",
      "Test accuracy: 0.99609375\n"
     ]
    }
   ],
   "source": [
    "#predicate\n",
    "score = model.evaluate(test_generator, steps=24,verbose=1)\n",
    "print('Test score:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "eb75df2a-897e-4388-82b3-3aa44c71688f",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}