Openpose: faq out of memory clarification regarding Windows demo
1180 raised the issue, but the faqs are still unclear about whether you need to install/enable cuDNN when using the Windows demo. Does enabling cuDNN need to be done with both the demo and when building from source?
kjacks21
All 2 comments
You are right, I'll update the faq. This will be the updated one. Thanks for the question!
Out of Memory Error
Q: Out of memory error - I get an error similar to: Check failed: error == cudaSuccess (2 vs. 0) out of memory.
A: Make sure you have a GPU with at least 4 GB of memory. If your GPU is between 2 and 4 GB, it should be fine for body-only settings, but you can also reduce the --net_resolution if it does not fit (check Speed Up, Memory Reduction, and Benchmark for the consequences of this).
(Only if you are compiling OpenPose by yourself, this does not apply to the portable OpenPose binaries for Windows because they already include cuDNN): If you have a GPU with >= 4 GB of memory, and you still face this error, most probably cuDNN is not installed/enabled. The default Caffe model uses >12 GB of GPU memory, cuDNN reduces it to ~2.2 GB for BODY_25 (default) and ~1.5 GB for COCO (--model_pose COCO). Note that you still need at least about 2.2 GB free for the default OpenPose to run. I.e., GPUs with only 2 GB will not fit the default OpenPose, and you will have to either switch to the COCO model (slower and less accurate), or reduce the --net_resolution (faster speed but also lower accuracy).
Also, hands and face increases the GPU memory requeriments, and 4 GB GPUs might run a bit short in some cases.
gineshidalgo99
on 29 Sep 2019
This resolution works for me " --net_resolution 320x320" with a decent performance.
sypei
on 18 Nov 2020
Related issues
shiangyoung
·
5Comments
MyraBaba
·
4Comments
ran2207
·
3Comments
anonymous530
·
4Comments
azuic
·
4Comments
Most helpful comment
You are right, I'll update the faq. This will be the updated one. Thanks for the question!
Out of Memory Error
Q: Out of memory error - I get an error similar to:
Check failed: error == cudaSuccess (2 vs. 0) out of memory.A: Make sure you have a GPU with at least 4 GB of memory. If your GPU is between 2 and 4 GB, it should be fine for body-only settings, but you can also reduce the
--net_resolutionif it does not fit (check Speed Up, Memory Reduction, and Benchmark for the consequences of this).(Only if you are compiling OpenPose by yourself, this does not apply to the portable OpenPose binaries for Windows because they already include cuDNN): If you have a GPU with >= 4 GB of memory, and you still face this error, most probably cuDNN is not installed/enabled. The default Caffe model uses >12 GB of GPU memory, cuDNN reduces it to ~2.2 GB for BODY_25 (default) and ~1.5 GB for COCO (
--model_pose COCO). Note that you still need at least about 2.2 GB free for the default OpenPose to run. I.e., GPUs with only 2 GB will not fit the default OpenPose, and you will have to either switch to theCOCOmodel (slower and less accurate), or reduce the--net_resolution(faster speed but also lower accuracy).Also, hands and face increases the GPU memory requeriments, and 4 GB GPUs might run a bit short in some cases.