Non-Max-Suppression

[EddieEduardo]

I'm still confused about one place in utils.py/non_max_suppression

Line244-Line246 "" score = image_pred[ : , 4] * image_pred[ : , 5 :].max(1)[0]

Sore by it

image_pred = image_pred[ (-score).argsort() ] "" The place I'm confused about is [0] after max(1) , why it has to be like this way rather than "" score = image_pred[ : , 4] * image_pred[ : , 5 :].max(1) ""

[pandalgx]

tensor.max(1) will return the score and its corresponding index, together. Here we just want to get the score, which has the index of 0.

[EddieEduardo]

Thanks，I mistook it as the first score value of a tensor.