Ubuntu Native NNStreamer Application Example - Pose Estimation (Single Person)

Introduction

This example passes camera video stream to a neural network using tensor_filter. The neural network predicts positions of 17 body parts of the person in input stream. The results are drawn by cairooverlay GStreamer plugin.

How to Run

This example requires pose estimation tf lite model and key point label (for body parts).
The public model can be obtained from this link: https://storage.googleapis.com/download.tensorflow.org/models/tflite/posenet_mobilenet_v1_100_257x257_multi_kpt_stripped.tflite
get-model.sh download the required model and make label text file

# bash
$ cd $NNST_ROOT/bin
$ ./get-model.sh pose-estimation-tflite
$ export GST_PLUGIN_PATH=$GST_PLUGIN_PATH:$NNST_ROOT/lib/gstreamer-1.0
$ ./nnstreamer_example_pose_estimation_tflite

Screenshots