Tong He comments

Results 20 comments of


                                            Tong He

Errors when compiling caffe.

Can you lower grade your gcc version. I was using 4.8.4

This is all the link for building linux-vdso.so.1 => (0x00007ffca0212000) libcudart.so.8.0 => /usr/local/cuda-8.0/lib64/libcudart.so.8.0 (0x00007fc42275d000) libcublas.so.8.0 => /usr/local/cuda-8.0/lib64/libcublas.so.8.0 (0x00007fc41fcc5000) libcurand.so.8.0 => /usr/local/cuda-8.0/lib64/libcurand.so.8.0 (0x00007fc41bd4f000) libglog.so.0 => /usr/lib/x86_64-linux-gnu/libglog.so.0 (0x00007fc41baf3000) libgflags.so.2 => /usr/lib/x86_64-linux-gnu/libgflags.so.2 (0x00007fc41b8d3000)...

Errors when compiling caffe.

@DreadPiratePsyopus I think this is easy to solve. You can transfer the custom c++ layer to any version caffe you use.

Can't read test_iou.pt but test_lstm.pt works

@Bshowg May be there exist some layers that don't exist in caffe mobile-lib

Cannot make caffe version due to reverse_axis_layer error

@DreadPiratePsyopus, Thanks for your interest. I have rebuild it but can't replicate this error. I think you can have a try to use CPU version of this layer.

@tonghe90 我这边没有12gb的显卡，可否用两块6gb的显卡（比如GTX1060）替代？

@wenston2006 这个我没有做过实验，不确定batchsize 对performance的影响

Sometimes proposal_idx was empty

Hi, please check https://github.com/aim-uofa/DyCo3D/issues/8#issuecomment-888110156.

Sometimes proposal_idx was empty

That depends on the model you use. If the channel unit (the value of m in the config file) is 16, 1080TI (11G) should be good

Problem when reproducing the result

@PinTang1999 Thanks for your interest. I would recommend you to train a network with segmentation, offset prediction tasks. Then you can fine-tune this model.

Implementation detail on S3DIS dataset

We will release the code for S3DIS in the future