I am wondering if np.zeros should be used, in order to create an empty array, instead of np.ones.
Using np.ones to initialize cam could bias cam towards positive values. So the effects by cam = np.maximum(cam, 0)would be reduced.
I am wondering if I understood the algo correctly. Thank you again !
You can have a look at #11. It is funny that I said I should add a comment to the code to avoid confusion and still did not do it for about 3 years. I added that comment now. Thanks.
Hi there,
Many thanks for the demo. It is very helpful. I have a question here: Line 81 in gradcam.py:
I am wondering if np.zeros should be used, in order to create an empty array, instead of np.ones. Using np.ones to initialize cam could bias cam towards positive values. So the effects by
cam = np.maximum(cam, 0)
would be reduced.I am wondering if I understood the algo correctly. Thank you again !