@pjreddie
Hi
Can you please confirm that all yolo layers have mask values 0,1,2 multiple times at the end . Is it right or it should have different values in different yolo layers
ahsan856jalal
All 6 comments
oh shit sorry about that, fixed e4acba689c89e223636efe7dabe8f55ae019cd8a
pjreddie
on 26 Mar 2018
Thanks Mate !
ahsan856jalal
on 26 Mar 2018
One last question , in yolo v2 , we have to set last convolutiona layer's filters like (classes+1+coords)*num
e.g for coords=4, num=5 and classes =1, filters will be 30.
However, you have used filters to be 75 for classes=20 and num =9.
So what if I have 92 classes and five anchors ?
ahsan856jalal
on 26 Mar 2018
num is 9 but each yolo layer is only actually looking at 3 (that's what the mask thing does). so it's (20+1+4)*3 = 75. If you use a different number of anchors you have to figure out which layer you want to predict which anchors and the number of filters will depend on that distribution
pjreddie
on 26 Mar 2018
Thanks
ahsan856jalal
on 26 Mar 2018
@pjreddie mask means just detect anchor masked ?
hexiangquan
on 29 Mar 2018
Related issues
ghost
·
4Comments
bujingdexin
·
3Comments
arianaa30
·
3Comments
Vikalp-Reorder
·
3Comments
kthordarson
·
3Comments
Most helpful comment
num is 9 but each yolo layer is only actually looking at 3 (that's what the mask thing does). so it's (20+1+4)*3 = 75. If you use a different number of anchors you have to figure out which layer you want to predict which anchors and the number of filters will depend on that distribution