rtmlib 0.0.9

RTMPose series models without mmcv, mmpose, mmdet etc.

rtmlib

rtmlib is a super lightweight library to conduct pose estimation based on RTMPose models WITHOUT any dependencies like mmcv, mmpose, mmdet, etc.

Basically, rtmlib only requires these dependencies:

• numpy
• opencv-python
• opencv-contrib-python
• onnxruntime

Optionally, you can use other common backends like opencv, onnxruntime, openvino, tensorrt to accelerate the inference process.

• For openvino users, please add the path <your python path>\envs\<your env name>\Lib\site-packages\openvino\libs into your environment path.

Installation

• install from pypi:

pip install rtmlib -i https://pypi.org/simple

• install from source code:

git clone https://github.com/Tau-J/rtmlib.git
cd rtmlib

pip install -r requirements.txt

pip install -e .

# [optional]
# pip install onnxruntime-gpu
# pip install openvino

Quick Start

Run webui.py:

# Please make sure you have installed gradio
# pip install gradio

python webui.py

Here is also a simple demo to show how to use rtmlib to conduct pose estimation on a single image.

import cv2

from rtmlib import Wholebody, draw_skeleton

device = 'cpu'  # cpu, cuda
backend = 'onnxruntime'  # opencv, onnxruntime, openvino
img = cv2.imread('./demo.jpg')

openpose_skeleton = False  # True for openpose-style, False for mmpose-style

wholebody = Wholebody(to_openpose=openpose_skeleton,
                      mode='balanced',  # 'performance', 'lightweight', 'balanced'. Default: 'balanced'
                      backend=backend, device=device)

keypoints, scores = wholebody(img)

# visualize

# if you want to use black background instead of original image,
# img_show = np.zeros(img_show.shape, dtype=np.uint8)

img_show = draw_skeleton(img_show, keypoints, scores, kpt_thr=0.5)


cv2.imshow('img', img_show)
cv2.waitKey()

APIs

• Solutions (High-level APIs)
  • Wholebody
  • Body
  • PoseTracker
• Models (Low-level APIs)
  • YOLOX
  • RTMDet
  • RTMPose
    • RTMPose for 17 keypoints
    • RTMW for 133 keypoints
    • RTMO for one-stage pose estimation (17 keypoints)
• Visualization
  • draw_bbox
  • draw_skeleton

TODO

• Support MMPose-style skeleton visualization
• Support OpenPose-style skeleton visualization
• Support WholeBody
• Support Hand
• Support Face
• Support Animal
• Support ONNXRuntime backend
• Support auto download and cache models
• Lightweight models
• Support 3 modes: performance, lightweight, balanced to select
• Support alias to choose model
• Support naive PoseTracker
• Support OpenVINO backend
• Support TensorRT backend
• Gradio interface
• Compatible with Controlnet
• Support RTMO

Model Zoo

By defaults, rtmlib will automatically download and apply models with the best performance. But you can also specify the model you want to use by passing the onnx_model argument.

More models can be found in RTMPose Model Zoo.

Detectors

Person

Notes:

• Models trained on HumanArt can detect both real human and cartoon characters.
• Models trained on COCO can only detect real human.

ONNX Model	Input Size	Description
YOLOX-l	640x640	trained on COCO val2017
YOLOX-nano	416x416	trained on HumanArt
YOLOX-tiny	416x416	trained on HumanArt
YOLOX-s	640x640	trained on HumanArt
YOLOX-m	640x640	trained on HumanArt
YOLOX-l	640x640	trained on HumanArt
YOLOX-x	640x640	trained on HumanArt

Pose Estimators

Body

ONNX Model	Input Size	Description
RTMPose-t	256x192	Body 17 Keypoints
RTMPose-s	256x192	Body 17 Keypoints
RTMPose-m	256x192	Body 17 Keypoints
RTMPose-l	384x288	Body 17 Keypoints
RTMPose-x	384x288	Body 17 Keypoints
RTMO-s	640x640	Body 17 Keypoints
RTMO-m	640x640	Body 17 Keypoints
RTMO-l	640x640	Body 17 Keypoints

WholeBody

ONNX Model	Input Size	Description
RTMW-m	256x192	Wholebody 133 Keypoints
RTMW-l	256x192	Wholebody 133 Keypoints
RTMW-l	384x288	Wholebody 133 Keypoints
RTMW-x	384x288	Wholebody 133 Keypoints

Visualization

MMPose-style	OpenPose-style

Citation

@misc{rtmlib,
  title={rtmlib},
  author={Jiang, Tao},
  year={2023},
  howpublished = {\url{https://github.com/Tau-J/rtmlib}},
}

@misc{jiang2023,
  doi = {10.48550/ARXIV.2303.07399},
  url = {https://arxiv.org/abs/2303.07399},
  author = {Jiang, Tao and Lu, Peng and Zhang, Li and Ma, Ningsheng and Han, Rui and Lyu, Chengqi and Li, Yining and Chen, Kai},
  keywords = {Computer Vision and Pattern Recognition (cs.CV), FOS: Computer and information sciences, FOS: Computer and information sciences},
  title = {RTMPose: Real-Time Multi-Person Pose Estimation based on MMPose},
  publisher = {arXiv},
  year = {2023},
  copyright = {Creative Commons Attribution 4.0 International}
}

@misc{lu2023rtmo,
      title={{RTMO}: Towards High-Performance One-Stage Real-Time Multi-Person Pose Estimation},
      author={Peng Lu and Tao Jiang and Yining Li and Xiangtai Li and Kai Chen and Wenming Yang},
      year={2023},
      eprint={2312.07526},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Acknowledgement

Our code is based on these repos:

• MMPose
• RTMPose
• DWPose