Jianzhong He

Hi @adithyabenny , I use the code you commit, and still encounter the problem that cudaErrorIllegalAddress and the error message is CUDA error at lanczos.cu:217 code=77(cudaErrorIllegalAddress) "cudaMemcpy(devVector, d_aVectorQQ, nPixels *...

and here is the completely output ` ./bin/linux/release/damascene damascene/polynesia.ppm Using cuda device 2: GeForce GTX TITAN X Processing: damascene/polynesia.ppm, output in damascene/polynesiaPb.pgm and damascene/polynesia.pb Eig 9 Tol 0.001000 Texton 1...

Hi @gaopeng-eugene @qizhuli , do you have the initialization model, could you please share it to me, thanks.

@zhapeng you need uncomment the MATLAB_DIR in Makfile.convfig and assign it to the path of your matlab

Thank you for great work. what if i use horovod on a single gpu machine?I tried it and found it fast than not use horovod, do this have any problem?Moreover,...

Hi, @uulm-mschoen do you have got the updated version. I meet the same problem.

That file is only the list of test image, including two columns and they are same(just convenient to read for the dataset class), such as 'path/im path/im'

Sorry，I do not know what happened to you just according the description you give. I guess maybe the image is too large or the image path is not correct?

No, we don't stop the gradient of upsample. we just stop the gradient to the previous predictions which we suppose the prediction is the proper edge that layer predict.

Thanks for your attention of our work, the below is the groundtruth of NYUD. [groundtruth.zip](https://github.com/pkuCactus/BDCN/files/4948672/groundtruth.zip)