Related Experiment Video
Updated: Mar 2, 2026

07:34
Utilizing vmTracking to Improve the Accuracy of Multi-Animal Pose Estimation in Rodent Social Behavior Studies
Published on: November 7, 2025
364
Learning to recognize rat social behavior: Novel dataset and cross-dataset application.
Malte Lorbach1, Elisavet I Kyriakou2, Ronald Poppe3
1Department of Information and Computing Sciences, Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands; Noldus Information Technology BV, Nieuwe Kanaal 5, 6709 PA Wageningen, The Netherlands.
Journal of Neuroscience Methods
|May 13, 2017
Summary
We introduce RatSI, the first public rat social interaction dataset, to train and validate automated rodent behavior analysis. Cross-dataset validation is crucial for reliable machine learning models in different experimental settings.
Area of Science:
- Behavioral neuroscience
- Machine learning in animal behavior research
Background:
- Automated tools using machine learning are replacing manual annotation for rodent social behavior analysis.
- Training and validation require diverse experimental data for generalizability.
Purpose of the Study:
- Introduce RatSI, the first publicly available dataset for rat social interaction.
- Enable development and validation of automated rodent behavior recognition methods.
Main Methods:
- Developed and released the RatSI dataset.
- Trained a rat interaction recognition method using the dataset.
- Performed cross-dataset validation to assess model performance across different settings.
Main Results:
- Demonstrated the utility of RatSI for training interaction recognition models.
- Showcased performance degradation due to experimental setting variations.
- Improved model performance by incorporating a simple adaptation step.
Conclusions:
- RatSI facilitates the development of robust automated rodent behavior recognition.
- Cross-dataset validation is essential for understanding classifier performance.
- Adaptation techniques can enhance the applicability of automated methods across diverse settings.

