miau.py

    #prepare gradcam
    image = cv2.imread(args.image, 1) #1 color, 0 grayscale
    """
    warning:: ensure slicing of imagepath is set to the right dimension, otherwise cam generation will fail. Probably with "NoneType" warning. 
    param: img_args_offset: set offset to image args
    """
    #check image paths and enter offset
    img_args_offset = 54
    print(args.image[img_args_offset:-4])

    #image network preprocessing
    image = np.float32(cv2.resize(image, (HEIGHT, WIDTH)))
    original = image
    combined = image
    cv2.imwrite('./evaluation/cam_output/{}-original.png'.format(args.image[img_args_offset:-4]), image) #save original for later comparison
    image = preprocessing_function(image.reshape(1, HEIGHT, WIDTH, 3))
    
    #prepare layers for gradcam
    layer_idx = visutils.find_layer_idx(finetune_model, 'dense_2') #last dense layer 310, other forum post says 312 (dense_2 = 313, dense_1 = 312(?))
    print("layer_idx: {}".format(layer_idx))

    # use explainer module
    explainer = GradCam(finetune_model, layer=layer_idx)
    exp_heatmap = explainer.explain(image, None) #target class 744, returns array

    # We resize the heatmap to have the same size as the original image
    exp_heatmap = cv2.resize(exp_heatmap, (c_original.shape[1], c_original.shape[0]))

    # We convert the heatmap to RGB
    #exp_heatmap = np.uint8(255 * exp_heatmap)

    # We apply the heatmap to the original image
    #exp_heatmap = cv2.applyColorMap(heatmap, cv2.COLORMAP_JET)

    # 0.4 here is a heatmap intensity factor
    #superimposed_img = exp_heatmap * 0.4 + c_original

    # Save the image to disk
    cv2.imwrite('./evaluation/cam_output/{}-blended_exp.png'.format(args.image[img_args_offset:-4]), exp_heatmap)
    print("cam visualized")