Refactored to use data.utils train/val split function

notebooks/demo/tf_flowers.ipynb

@@ -67,27 +67,18 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "# get autoencoder image dataset tools\n",
+    "from autoencoder.data.utils import train_val_split_image_dataset\n",
+    "\n",
     "# use keras util to load raw images into tensorflow.data.Dataset\n",
-    "x_train, x_val = keras.utils.image_dataset_from_directory(\n",
+    "x_train, x_val = train_val_split_image_dataset(\n",
     "  data_dir,\n",
     "  color_mode=\"rgb\",\n",
     "  validation_split=(1 / 7),\n",
-    "  subset=\"both\",\n",
     "  seed=42,\n",
     "  image_size=(28, 28),\n",
     "  batch_size=32,\n",
-    ")\n",
-    "\n",
-    "# use a keras layer to normalize\n",
-    "normalization_layer = keras.layers.Rescaling(1./255)\n",
-    "\n",
-    "# will apply normalization as well as regrouping each element\n",
-    "def change_inputs(images, labels):\n",
-    "  x = normalization_layer(images)\n",
-    "  return x, x\n",
-    "\n",
-    "# apply change\n",
-    "x_train, x_val = x_train.map(change_inputs), x_val.map(change_inputs)"
+    ")"
    ]
   },
   {
@@ -217,7 +208,7 @@
     "ae_eval = AutoencoderEvaluator(autoencoder, x_val)\n",
     "\n",
     "# view distribution\n",
-    "ae_eval.view_error_distribution(\"tf_flowers Autoencoder: Reconstruction Error Distribution\")"
+    "ae_eval.view_error_distribution(\"tf_flowers Autoencoder: Reconstruction Error Distribution\", bins=100)"
    ]
   }
  ],