TensorRT Pose Estimation

This project features multi-instance pose estimation accelerated by NVIDIA TensorRT. It is ideal for applications where low latency is necessary. It includes

  • Training scripts to train on any keypoint task data in MSCOCO format

  • A collection of models that may be easily optimized with TensorRT using torch2trt


Human pose estimation


This task involves detecting human body pose using models trained on the MSCOCO dataset.


Below are models pre-trained on the MSCOCO dataset. The throughput in FPS is shown for each platform

Model Jetson Nano Jetson Xavier Weights
resnet18_baseline_att_224x224_A 22 251 download (81MB)
densenet121_baseline_att_256x256_B 12 101 download (84MB)

Live demo

To run the live Jupyter Notebook demo on real-time camera input, follow these steps

  1. Place the downloaded weights in the tasks/human_pose directory

  2. Open and follow the live_demo.ipynb notebook

    You may need to modify the notebook, depending on which model you use


To install trt_pose, call this command

We assume you have already installed PyTorch, torchvision, and TensorRT

sudo pip3 install tqdm cython pycocotools
sudo apt-get install python3-matplotlib
git clone https://github.com/NVIDIA-AI-IOT/trt_pose
cd trt_pose
sudo python3 setup.py install

See also

  • torch2trt - An easy to use PyTorch to TensorRT converter

  • JetBot - An educational AI robot based on NVIDIA Jetson Nano

  • JetRacer - An educational AI racecar using NVIDIA Jetson Nano

  • JetCam - An easy to use Python camera interface for NVIDIA Jetson


Cao, Zhe, et al. "Realtime multi-person 2d pose estimation using part affinity fields." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017.

Xiao, Bin, Haiping Wu, and Yichen Wei. "Simple baselines for human pose estimation and tracking." Proceedings of the European Conference on Computer Vision (ECCV). 2018.