BodyPoseNet in Deepstream

[edit_or]

DeepStream5.1 current version doesn’t have official support for this BodyPoseNet application.
But there is one sample app for DeepStream SDK.

Can I use this BodyPoseNet with this sample app?

Can I use directly with this app or any changes I need to make to use this BodyPoseNet in DeepStream?
DeepStream6.0 will support BodyPoseNet, but not sure when it will be launched officially. I apply for early access but not approved yet.

This pose explain for training BodyPose , is it possible to change other network like Resnet instead of using VGG?
How can I do that?

[Morganh]

Currently, only the vgg19 backbone is supported. Its pretrained model locates at https://ngc.nvidia.com/catalog/models/nvidia:tlt_bodyposenet/files

Please refer to
https://docs.nvidia.com/tlt/tlt-user-guide/text/bodypose_estimation/bodyposenet.html#deploying-to-the-tlt-cv-inference-pipeline to deploy the bodypose model.

You can deploy a model trained through the TLT workflow to the TLT CV Inference Pipeline. The TLT CV Quick Start Scripts section contains instructions for both options.

More info for bodypose, please refer to https://docs.nvidia.com/tlt/tlt-user-guide/text/bodypose_estimation/index.html
and blogs https://developer.nvidia.com/blog/training-optimizing-2d-pose-estimation-model-with-tlt-part-1 , https://developer.nvidia.com/blog/training-optimizing-2d-pose-estimation-model-with-tlt-part-2/

[edit_or]

How can I use this BodyPoseNet in DeepStream?

[Morganh]

You can apply for DS6.0 early access.
https://developer.nvidia.com/deepstream-sdk-6.0-early-access

• New AI models such as ASR, pose estimation, facial landmark and emotion expand the use-cases that can be easily deployed with DeepStream.