Algorithm 1: Creating and Training the Hybrid Segmentation-Classification Model

Require: Input Shape = (256, 256, 3)    1: Define Segmentation Model    2: inputs ← Input(input_shape)    3: conv1 ← Conv2D(32, (3, 3), activation=`relu’, padding=`same’)(inputs)    4: pool1 ← MaxPooling2D(pool_size=(2, 2))(conv1)    5: up1 ← UpSampling2D(size=(2, 2))(pool1)    6: decoded ← Conv2D(3, (3, 3), activation=`sigmoid’, padding=`same’)(up1)    7: segmentation_model ← Model(inputs, decoded, name=`segmentation_model’)    8: Define Classification Model    9: inputs ← Input(input_shape)  10: conv1 ← Conv2D(16, (3, 3), activation=`relu’)(inputs)  11: pool1 ← MaxPooling2D(pool_size=(2, 2))(conv1)  12: conv2 ← Conv2D(32, (3, 3), activation=`relu’)(pool1)  13: pool2 ← MaxPooling2D(pool_size=(2, 2))(conv2)  14: conv3 ← Conv2D(64, (3, 3), activation=`relu’)(pool2)  15: pool3 ← MaxPooling2D(pool_size=(2, 2))(conv3)  16: flatten ← Flatten()(pool3)  17: dense1 ← Dense(256, activation=`relu’)(flatten)  18: output ← Dense(1, activation=`sigmoid’)(dense1)  19: Combine Segmentation and Classification Models  20: input_tensor ← Input(input_shape)  21: segmentation_model ← create_segmentation_model()  22: segmentation_output ← segmentation_model(input_tensor)  23: classification_model ← create_classification_model()  24: classification_output ← classification_model(segmentation_output)  25: hybrid_model ← Model(inputs=input_tensor, outputs=classification_output, name=`hybrid_model’)  26: Compile and Train the Hybrid Model  27: hybrid_model.compile(optimizer=`adam’, loss=`binary_crossentropy’, metrics=[`acc’])  28: logdir=`logs’  29: hist ← hybrid_model.fit(train, epochs=25, validation_data=val, callbacks=[tensorboard_callback])