Which color map is returned in example script cam.py ?

[CristianCosci]

In the example file cam.py there is the following code:

with cam_algorithm (model = model,
                      target_layers = target_layers,
                      use_cuda = args.use_cuda) as cam:

       # AblationCAM and ScoreCAM have batch implementations.
       # You can override the internal batch size for faster calculation.
       cam.batch_size = 32
       grayscale_cam = cam (input_tensor = input_tensor,
                           goals = goals,
                           aug_smooth = args.aug_smooth,
                           eigen_smooth = args.eigen_smooth)

       # Here grayscale_cam has only one image in the batch
       grayscale_cam = grayscale_cam [0,:]

       cam_image = show_cam_on_image (rgb_img, grayscale_cam, use_rgb = True)

       # cam_image is RGB encoded while "cv2.imwrite" requires BGR encoding.
       cam_image = cv2.cvtColor (cam_image, cv2.COLOR_RGB2BGR)

My question was about the format of the grayscale_cam variable. as I noticed that the values ​​are in the range 0-1 and not 0-255 despite being (I think) a grayscale color map.
Having to use this variable to get a mask, is it correct if I carry the range between 0-255 simply by doing grayscale_cam * 255 ?

[CristianCosci]

And for example if i try to use an adaptiveThreshold like this:

img_thresh_adp = cv2.adaptiveThreshold(grayscale_cam, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 13, 7)

i occur this error

---------------------------------------------------------------------------
error                                     Traceback (most recent call last)
[<ipython-input-10-eb93012ef834>](https://localhost:8080/#) in <module>
      6 
      7 # Perform adaptive thresholding.
----> 8 img_thresh_adp = cv2.adaptiveThreshold(grayscale_cam, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 13, 7)
      9 
     10 # Display the images.

error: OpenCV(4.6.0) /io/opencv/modules/imgproc/src/thresh.cpp:1674: error: (-215:Assertion failed) src.type() == CV_8UC1 in function 'adaptiveThreshold'

There is some error in the type of grayscale_cam and i don't know what.
But if i save the image and reload it everything its ok.

cv2.imwrite('Mask.png', img_thresh);
mask = cv2.imread("Mask.png", cv2.IMREAD_GRAYSCALE)

Can anyone help me? I think it's not the best way to save and reload the image.

[jacobgil]

grayscale_cam is a float32 image with values in the range [0, 1].
Just convert it to np.uint8:
grayscale_cam_uint8 = np.uint8(grayscale_cam * 255)
and save grayscale_cam.

[CristianCosci]

Thanks so much