Pose Estimation on Deepstream

There is a example code of Pose Estimation on Deepstream ?

Thanks!

Hi, I want the same thing, until now i have found this repo and I am trying to use it with python binding.

If you know another alternative let us know, thx

2 Likes

DS does not support it because there is not post-processing for it in DS.

1 Like

Hi @joaquin1,
Sorry! The latest update is we will open source DeepStream post-processing for trt-pose by the end of this month.

2 Likes

Great thanks! Is any way i can get notify about this release

1 Like

We will normally notify the release with a pinned topic in the top of this forum,

2 Likes

@mchi it is the end of the month, any update on this. I know many people are eager for this to come to open source.

2 Likes

Hi,
Checked internally, it was delayed, but will have update in these weeks.

Sorry for the delay!

2 Likes

Hey everyone!

This code was released recently. See if this helps you out at all!

4 Likes

Looking good! Thanks for the update.

Cool! Thank you for the update!

hello @anujsaharan, thank you for your great work, but I try to run but failure cause osd library need libcudart.so.11.0. Could you provide osd build in cuda 10.2 library please!

Doesn’t the performance decrease if we squeeze a 16:9 video to the 1:1 NN? Is there anything to make the NN input/output shapes dynamic?

Hi,

The binary is available for cuda10.2 now. Please check: https://github.com/NVIDIA-AI-IOT/deepstream_pose_estimation/blob/master/bin/x86

Thanks!

1 Like

Dear my friend, I am running this deepstream-pose project in the docker nvcr.io/nvidia/deepstream:5.0-dp-20.04-devel , and the binary file libnvds_osd.so works not good.

“undefined symbol: nvds_mask_utils_resize_to_binary_argb32”

the CUDA nvcc in docker is 10.2

Hi zhujiandream,

Please help to open new topic if it’s still an issue. Thanks