search

qqwweee / keras-yolo3

A Keras implementation of YOLOv3 (Tensorflow backend)
MIT License
7.14k stars 3.45k forks source link

Consistent anchor mask between normal and tiny version #315

Open pprados opened 5 years ago

pprados commented 5 years ago

Hello,

I am not sure to understand all the code, but in the source code, we can find some line like this: anchor_mask = [[6, 7, 8], [3, 4, 5], [0, 1, 2]] if num_layers == 3 else [[3, 4, 5], [1, 2, 3]] Why is not this ? anchor_mask = [[6, 7, 8], [3, 4, 5], [0, 1, 2]] if num_layers == 3 else [[3, 4, 5], [0, 1, 2]] Is it normal or a bug ?

Thanks

xiaomaxiao commented 5 years ago

i think it's a bug

Frank-Jie commented 5 years ago

i also find this part when i use tiny-yolo , and i think it's a bug. but hope someone can point out if it really have some reason.

Truantboy commented 5 years ago

I tried [0,1,2] and found that the result seems worse than [1,2,3]. It may not be a bug but I can not find the reason.