caffe-fast-rcnn icon indicating copy to clipboard operation
caffe-fast-rcnn copied to clipboard

Published 20 hours ago • verified publisher icon rbgirshick

CPU version of smooth L1 loss layer implemented based

Open patricio-astudillo opened this issue 8 years ago • 2 comments

CPU version of smooth L1 loss layer implemented based on the GPU version. Peer review requested.

patricio-astudillo avatar Jul 07 '16 08:07 patricio-astudillo

@patricio-astudillo @rbgirshick I implemented the code based on GPU code of smooth l1 loss, but the gpu code doesnt take weights into consideration in its back propagation. And the back propagation in this CPU implementation includes it. Is this a mistake or am I missing something?

saiprabhakar avatar Aug 16 '16 17:08 saiprabhakar

How to use the CPU layer of SmoothL1Loss Layer? Just copy your code to the original one?

Lukeeeeee avatar Dec 13 '16 13:12 Lukeeeeee