Welcome to the Social-Seq Project

These are some explanatory documents for the Social-Seq project. Social-seq is used to analyze the 3D poses, social categories, and emotional states behind social interactions of animals in close contact. By constructing a real-time behavior classification system, it is potentially useful for treating autism (ASD) social disorders.

  • Code Repository 📦：https://github.com/LiLab-CIBR/Social_Seq

  • Document here 📚：https://lilab-cibr.github.io/Social_Seq/

Abstract Diagram


Contents

  • Hardware and Software System Installation - Install multi-camera systems, debug OBS, and enable F2 synchronization assistant.
  • Ball Calibration - Calibrate the spatial coordinate system of the multi-camera system.
  • Mouse Image Segmentation - Use segmentation models to enhance the signal-to-noise ratio of mice in complex backgrounds. Single mouse.
  • Social 3D Keypoint Reconstruction - Two-rat segmentation, keypoint prediction, and smoothing.
  • Seq Behavior Labeling - Obtain 36 categories of consistent social behavior labels for rats.

Code Release 📅

Last updated on 2025-8-15 by ChenXinfeng. 👨‍💻

References 📚

Xinfeng Chen; Xianming Tao; Zhenchao Zhong; Yuanqing Zhang; Yixuan Li; Ye Ouyang; Zhaoyi Ding; Min An; Miao Wang; Ying Li* (2025). Decoding the Valence of Developmental Social Behavior: Dopamine Governs Social Motivation Deficits in Autism. In preparation.

陈昕枫. (2025). 基于深度学习的动物自由社交行为分析研究. 博士论文，北京大学.

Corresponding Author 📬

  • Li Ying: liying