Robust mouse tracking in complex environments using neural networks.

Area of Science:

• Biobehavioral research
• Computational neuroscience
• Animal behavior analysis

Background:

• Accurate animal tracking is crucial for behavioral experiments.
• Current video-based methods often require environmental manipulation for segmentation, limiting scalability and ethological relevance.
• There is a need for robust tracking methods that accommodate long-term monitoring, dynamic environments, and heterogeneous animal populations.

Purpose of the Study:

• To evaluate state-of-the-art neural network architectures for single mouse tracking.
• To compare the performance of different neural network models under diverse visual and environmental conditions.
• To provide accessible tools and resources for the biobehavioral research community.

Main Methods:

• Application of a state-of-the-art neural network-based tracker for automated animal tracking.
• Comparative analysis of three distinct neural network architectures (e.g., encoder-decoder segmentation networks).
• Evaluation across visually diverse mice and varying environmental conditions.

Main Results:

• An encoder-decoder segmentation neural network demonstrated high accuracy and speed in tracking.
• The chosen network architecture required minimal labeled training data for effective performance.
• The study identified optimal hyperparameters for the neural network tracker.

Conclusions:

• Neural network-based tracking offers a robust solution for automated animal behavior analysis.
• Encoder-decoder segmentation networks are particularly effective for mouse tracking, even with limited training data.
• The developed tools and resources will aid researchers in long-term, dynamic behavioral monitoring.