Iamanorange

https://github.com/sampepose/flownet2-tf/issues/28#issuecomment-441118569

Search carefully before file an issue.

https://github.com/sampepose/flownet2-tf/issues/66#issuecomment-443215217

If you still get error: `expression must have a constant value`, it has no relevance with CUDA. If you get error: `identifier "__shfl" is undefined`, I have no idea. Both...

Clip the video to frames and run flownet2 for each pair of frame.

https://github.com/sampepose/flownet2-tf/issues/63#issuecomment-441422550

It doesn't matter wheather using conda or not.

Did you build these custom OPs on Windows? Failed when building OPs or running Flownet? What's the error log? I don't know if it can work on Windows. But the...

Try to add '-m' argument to your python interpreter. For PyCharm, open [Run/Debug Configurations], add '-m' (without quote) to [Interpreter Options].

@shoutashi @alisaaalehi @dehaisea For error "correlation.so: undefined symbol:_ZTIN10tensorflow8OpKernelE": Modify Makefile: **TF_LIB = \`python -c "import tensorflow; print(tensorflow.sysconfig.get_lib())"\`** CGPUFLAGS = -L$(CUDA_HOME)/lib -L$(CUDA_HOME)/lib64 -lcudart **-L$(TF_LIB) -ltensorflow_framework** If tensorflow.sysconfig.get_lib() cannot get correct dir,...