DeepLabCut/DeepLabCut-Backbones

DeepLabCut - Model Backbones

This repository contains backbone weights for DeepLabCut models [1]. These weights are downloaded automatically in DeepLabCut when a model architecture requiring them is used.

Backbone Architectures

CSPNeXt

The CSPNeXt backbone was first introduced in RTMDet: An Empirical Study of Designing Real-Time Object Detectors [2], and then used in RTMPose: Real-Time Multi-Person Pose Estimation based on MMPose [3]. These model weights are adapted from the CSPNeXt weights pre-trained on 7 public human pose estimation benchmarks released with the RTMPose models. Available variants are CSPNeXT-s, CSPNeXT-m, and CSPNeXT-x.

References

1. Alexander Mathis, Pranav Mamidanna, Kevin M. Cury, Taiga Abe, Venkatesh N. Murthy, Mackenzie W. Mathis, Matthias Bethge. DeepLabCut: markerless pose estimation of user-defined body parts with deep learning. In Nature Neuroscience, 21, 1281–1289 (2018).
2. Chengqi Lyu, Wenwei Zhang, Haian Huang, Yue Zhou, Yudong Wang, Yanyi Liu, Shilong Zhang, Kai Chen. RTMDet: An Empirical Study of Designing Real-Time Object Detectors. ArXiv, abs/2212.07784, 2022.
3. Tao Jiang, Peng Lu, Li Zhang, Ningsheng Ma, Rui Han, Chengqi Lyu, Yining Li, Kai Chen. RTMPose: Real-Time Multi-Person Pose Estimation based on MMPose. ArXiv, abs/2303.07399, 2023.