All_Inclusive

This contains the training files, models and implementation with device cameras. Also included is an attempt to convert to an app.

.gitattributes

@@ -1,2 +1,4 @@
 # Auto detect text files and perform LF normalization
 * text=auto
+*.pkg filter=lfs diff=lfs merge=lfs -text
+*.exe filter=lfs diff=lfs merge=lfs -text

Camera_tflite.py

@@ -0,0 +1,60 @@
+from kivy.app import App
+from kivy.uix.boxlayout import BoxLayout
+from kivy.uix.camera import Camera
+from kivy.uix.label import Label
+from kivy.graphics.texture import Texture
+import tensorflow as tf
+import tensorflow_hub as hub
+from kivy.clock import Clock
+import cv2
+import numpy as np
+
+class MyCameraApp(App):
+    def build(self):
+        self.layout = BoxLayout(orientation='vertical')
+        self.camera = Camera(play=True, resolution=(500, 500))
+        self.label = Label(text='Emotion will be displayed here')
+        self.layout.add_widget(self.camera)
+        self.layout.add_widget(self.label)
+        return self.layout
+
+    # def on_start(self):
+    #     # Load the TensorFlow Hub module
+    #     module_handle = 'https://www.kaggle.com/models/google/mobilenet-v2/frameworks/TensorFlow2/variations/035-224-feature-vector/versions/2'
+    #     module = hub.load(module_handle)
+    #     with hub.custom_object_scope({'KerasLayer': hub.KerasLayer(module)}):
+    #         self.model = tf.keras.models.load_model('Face (1).h5', custom_objects=custom_objects)
+    #     Clock.schedule_interval(self.update, 1.0 / 30.0)  # Update at 30 FPS
+
+    def on_start(self):
+    # Load the TensorFlow Hub module
+        module_handle = 'https://www.kaggle.com/models/google/mobilenet-v2/frameworks/TensorFlow2/variations/035-224-feature-vector/versions/2'
+        self.model = tf.keras.models.load_model('Face.h5', custom_objects={'KerasLayer': hub.KerasLayer})
+        Clock.schedule_interval(self.update, 1.0 / 30.0)  # Update at 30 FPS
+
+
+    def update(self, *args):
+        # Capture image from camera
+        camera_texture = self.camera.texture
+        camera_image_array = self.texture_to_array(camera_texture)
+        emotion = self.predict_emotion(camera_image_array)
+        self.label.text = f'Emotion: {emotion}'
+
+    def texture_to_array(self, texture):
+        buffer = texture.pixels
+        image_array = np.frombuffer(buffer, dtype=np.uint8)
+        image_array = image_array.reshape(texture.height, texture.width, 4)
+        image_array = cv2.cvtColor(image_array, cv2.COLOR_RGBA2RGB)
+        image_array = cv2.resize(image_array, (224, 224))  # Resize as per model input
+        image_array = image_array / 255.0  # Normalize
+        return image_array
+
+    def predict_emotion(self, image_array):
+        image_array = np.expand_dims(image_array, axis=0)
+        prediction = self.model.predict(image_array)
+        predicted_class = np.argmax(prediction, axis=1)
+        emotions = ["angry", "disgust", "fear", "happy", "neutral", "sad", "surprise"]
+        return emotions[predicted_class[0]]
+
+if __name__ == '__main__':
+    MyCameraApp().run()

Convert_Keras_tflite.py

@@ -0,0 +1,18 @@
+import tensorflow as tf
+import tensorflow_hub as hub
+
+# Specify the custom object
+custom_objects = {'KerasLayer': hub.KerasLayer}
+
+# Load your model with the custom object
+model = tf.keras.models.load_model('Face.keras', custom_objects=custom_objects)
+
+# Convert the model to TFLite
+converter = tf.lite.TFLiteConverter.from_keras_model(model)
+# Enable TF Select Ops
+converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS, tf.lite.OpsSet.SELECT_TF_OPS]
+tflite_model = converter.convert()
+
+# Save the TFLite model to a file
+with open('D:/Facial/archive (2)/Face.tflite', 'wb') as f:
+    f.write(tflite_model)

FED.py

@@ -0,0 +1,71 @@
+import tensorflow as tf
+import cv2
+import os
+import matplotlib.pyplot as plt
+import numpy as np
+import random
+from tensorflow import keras
+from tensorflow.keras import layers
+from tensorflow.keras.layers import GlobalAveragePooling2D, Dense
+from tensorflow.keras.models import Model
+
+'''
+img_array=cv2.imread("train/angry/Training_3908.jpg")
+
+plt.imshow(img_array)
+'''
+training = "D:/Facial/archive (2)/train" # The training directory
+
+classes = ["angry", "disgust", "fear", "happy", "neutral", "sad", "surprise"]
+
+for category in classes:
+    path = os.path.join(training, category)
+    for img in os.listdir(path):
+        img_array= cv2.imread(os.path.join(path,img))
+
+
+img_resize= 224
+resized_array= cv2.resize(img_array, (img_resize,img_resize)) 
+plt.imshow(cv2.cvtColor(resized_array, cv2.COLOR_BGR2RGB))
+plt.show()
+
+training_data = []
+
+def create_training_data():
+    for category in classes:
+        path = os.path.join(training, category)
+        class_num = classes.index(category)
+        for img in os.listdir(path):
+            try:
+                img_array= cv2.imread(os.path.join(path,img))
+                resized_array = cv2.resize(img_array, (img_resize,img_resize))
+                training_data.append([resized_array, class_num])
+            except Exception as e:
+                pass
+create_training_data()
+
+random.shuffle(training_data)
+
+X=[]
+y=[]
+
+for features, label in training_data:
+    X.append(features)
+    y.append(label)
+
+X= np.array(X).reshape(-1, img_resize, img_resize, 3)
+
+X=X/255.0 
+
+base_model = tf.keras.applications.MobileNetV2()
+
+base_input = base_model.layers[0].input
+base_output = base_model.layers[-2].output
+
+final_output= layers.Dense(128)(base_output)
+final_output = layers.Activation ('relu')(final_output)
+final_output = layers.Dense(64)(final_output)
+final_output = layers.Activation ('relu')(final_output)
+final_output = layers.Dense(7, activation = 'softmax')(final_output)
+
+final_model = keras.Model(inputs = base_input, outputs=final_output)