taosean

Results 12 comments of taosean

nvcc.hpp(22): fatal error C1017:

It seems commnet Line 22 to Line 24 in `nvcc.hpp` worked for me.

Convergence problem training with arc_torch/train_one_gpu.py

@eeric I used the script you provided to finetune the model glint360k_cosface_r100_fp16_0.1 downloaded [here](https://pan.baidu.com/s/1CL-l4zWqsI1oDuEEYVhj-g) on glint_asia dataset. My config is ``` config = edict() config.loss = "cosface" config.network = "r100"...

Convergence problem training with arc_torch/train_one_gpu.py

@eeric So you mean there maybe errors in the train_one_gpu.py? Do you have any idea where it is? Thanks!

Segmentation fault(Core Dump)？

@cwjhx @h72001346 Did you use gcc 4.9.2 or above? I met the same problem when compiling with gcc 4.8.5. I solved it by compiling with gcc 5.3.1. Hope this helps.

Annotation tool

I also want to know how to label my custom dataset like XFUND. Any help?

Annotation tool

Thanks, I'll have a look.

Waht's the architecture of the provided pretrained weight?

@fdbtrs ，thank you for your response. Since the provided pretrained models are IResnet100, do you have a plan to release the iresnet50 version pretrained models?

module use of python27.dll conflicts with this version of python

@anandkoirala Hi Anand, have you solved this problem, I almost had the same problem as yours.

module use of python27.dll conflicts with this version of python

@anandkoirala I am using python3.5 on Windows since tensorflow on Windows only supports python 3.5 or 3.6. This is a little bit annoying.

src/caffe/layers/cudnn_conv_layer.cpp:108:58:error:'cudnnSetConvolutionGroupCount' was not declared in this scope

When I use cuda8 cudnn5.1, I encounterd the same problem. Switching to cuda 9.0 cudnn 7.3 solved this problem for me.