Edgar Riba

We don't have grabcut. But still I see some false positives. The best would be detecting depth or large gradient intensity to filter out or add an attention loss

@wuhaozhe nice ! could you send a small PR to fix this ?

Hi, I got similar issues and I couldn't find the way to replicate the paper experiments and consequently I abandoned this project. If you find where is the issue, please...

/cc @szagoruyko @fmassa

@SevenGong thx for your interest in DeepRosetta ! Currently, the project is in a very early stage and all the codes are experimental. So do not expect to make it...

@SevenGong check last commit https://github.com/edgarriba/DeepRosetta/commit/4908092010c4e2fc4f9949b1f273c08996266b73

@Maratyszcza If you have time update reference to tiny-dnn (no tiny-cnn anymore!). Now NNPACK is fully integrated in master branch

@dprylipko also we support it in tiny-dnn https://github.com/tiny-dnn/tiny-dnn/blob/master/tiny_dnn/core/kernels/conv2d_op_nnpack.h

hey! thanks for your answer but I'm not sure if I understood it. Does it mean that you haven't test it? After running Torch models the resultant descriptors have size...

nice! I'm struggling a bit to get proper results with your network. In the following plot I've tried some feature extractors and injected the results to vlfeat benchmark in the...