keras-yolo3
keras-yolo3 copied to clipboard
Bounding boxes don't get eliminated when objectness is less than the threshold. Fix for a single line.
Hi @experiencor ,
There is a minor code that should be fixed when decoding the output of the network. the code is here: https://github.com/experiencor/keras-yolo3/blob/768c524f277adbfd26c2f44d73cb1826bbaf2d10/yolo3_one_file_to_detect_them_all.py#L302
I do not know a lot of python, however, your code doesn't continue when objectness is under the threshold. Because of this, it takes a lot of processing time for decoding bounding boxes because they don't get eliminated at the first step.
I think it should be something like this: