I saw 2 pages regarding pose_classification
In the first link, the Training data part says:
The model is trained on an NVIDIA dataset with 6 annotated action classes, i.e., sitting_down, getting_up, sitting, standing, walking and jumping. The skeletons are based on the 34-keypoint NVIDIA format generated by the deepstream-bodypose-3d app.
1.If I want to train a pose classification net using my own dataset using TAO 5.0, do I have to run deepstream-bodypose-3d to generate the keypoints representing body parts as the required training dataset? Not sure if I misunderstood anything.
2.I’m working on an AWS EC2 server, where one Tesla T4 GPU is available and TAO toolkit 5.0.0 has been installed so far and the installed CUDA version is 12.2. The OS is ubuntu 22.04, while the document page(link provided below) says that the components have to be installed:
- Ubuntu 20.04
- GStreamer 1.16.3
- NVIDIA driver 525.125.06
- CUDA 12.1
- TensorRT 220.127.116.11
libssl1.1 couldn’t be found to be installed by using sudo apt install command as it seems that it’s obsolete in Ubuntu 22.04.
Do I have to manually download and install it or use a newer version of libssl? Does the installation guide work for Ubuntu 22.04?