DeepOcclusion icon indicating copy to clipboard operation
DeepOcclusion copied to clipboard

Do you have caffe versions to implement the DeepOcclusion? since theano has been stopped developping

Open scut-salmon opened this issue 7 years ago • 8 comments

I have run successfully the RunParts.py but something go wrong with the RunUnaries.py, which use the ROI-pooling layers. theano.sanbox.cuda was not used and it went wrong with: from theano.sanbox.cuda import as_cuda_ndarray_variable, GpuOp still error when I use: from theano.gpuarray import as_cuda_ndarray_variable, GpuOp

how can I solve it ? Looking forward to your response!

scut-salmon avatar Feb 08 '18 05:02 scut-salmon

I have solved the problem and this project is work with: theano 0.8.2 cuDNN5.0 cuda8.0

scut-salmon avatar Feb 21 '18 04:02 scut-salmon

Can you share your solution please? I fixed the issues with GpuOp but the theano ROI pooling is not compiling with me

okoky-krypto avatar Feb 21 '18 13:02 okoky-krypto

Yes I also met some problem with RoI pooling files, and the reason is that I use theano version 1.01, which is not the suitable version. The code worked when I use theano 0.8.2 instead.

scut-salmon avatar Feb 25 '18 15:02 scut-salmon

@okoky-krypto did your issue get solved. If yes, please share the solution. Even am struck at ROI pooling compiling.

nagasanthoshp avatar May 04 '18 10:05 nagasanthoshp

@nagasanthoshp I actually could not solve the issue using theano 1.x and cuDNN7.0. However, when I reverted back to the mentioned versions by @scut-salmon I was able to run the code. I found out that if you use a small number of rectangles in your room setup you end up with an error because rois comes up empty. I do not know the reason for the error and I opened an issue for it but I did not investigate any further.

okoky-krypto avatar May 10 '18 10:05 okoky-krypto

@scut-salmon What version of pygpu did you use with theano 0.8.2?

MahajanKanika avatar Jul 12 '18 06:07 MahajanKanika

I have solved the problem and this project is work with: theano 0.8.2 cuDNN5.0 cuda8.0

I tried your versions of package and run the code, I got the error: ERROR (theano.sandbox.gpuarray): Could not initialize pygpu, support disabled Traceback (most recent call last): File "/home/fbh/anaconda3/envs/DO/lib/python2.7/site-packages/theano/sandbox/gpuarray/init.py", line 95, in init_dev(config.device) File "/home/fbh/anaconda3/envs/DO/lib/python2.7/site-packages/theano/sandbox/gpuarray/init.py", line 46, in init_dev "Make sure Theano and libgpuarray/pygpu " RuntimeError: ('Wrong major API version for gpuarray:', 1, 'Make sure Theano and libgpuarray/pygpu are in sync.') What causes this? Thank you very much

soullessrobot avatar Nov 20 '18 21:11 soullessrobot

@scut-salmon Where could I contact you? I've been having problems with this project and would appreciate a lot some help!

DiegoJRR avatar Dec 28 '19 04:12 DiegoJRR