m.fit(X, Y, verbose=1, epochs=100, shuffle=True)
# for e in range(20):
# goOn=input("Go on? y or n :")
# if goOn=='n':
# break
m.save_weights(MODEL_FILE_NAME)
# m.save(MODEL_FILE_NAME)
yPred = m.predict(X)
# yPred = yPred - np.min(yPred)
# yPred = yPred / np.max(yPred + EPS)
yPred = img_hndlr.inv_preprocess_images(yPred)
yPred = yPred.astype(np.uint8)
x1LayerOut = Model(xx, m.get_layer("X1").output).predict(X, verbose=1)
x1LayerOut = x1LayerOut - np.min(x1LayerOut)
x1LayerOut = x1LayerOut / np.max(x1LayerOut + EPS)
x1LayerOut = img_hndlr.inv_preprocess_images(x1LayerOut).astype(np.uint8)
x2LayerOut = Model(xx, m.get_layer("X2").output).predict(X, verbose=1)
x2LayerOut = x2LayerOut - np.min(x2LayerOut)
x2LayerOut = x2LayerOut / np.max(x2LayerOut + EPS)
x2LayerOut = img_hndlr.inv_preprocess_images(x2LayerOut).astype(np.uint8)
diffLayerOut = Model(xx, m.get_layer("DIFF").output).predict(X, verbose=1)
diffLayerOut = diffLayerOut - np.min(diffLayerOut)
diffLayerOut = diffLayerOut / np.max(diffLayerOut + EPS)
diffLayerOut = img_hndlr.inv_preprocess_images(diffLayerOut).astype(
np.uint8)
