Pose estimation and tracking dataset for multi-animal behavior analysis on the China Space Station

Area of Science:

• Animal neuroscience and ethology
• Artificial intelligence in behavioral analysis
• Space biology

Background:

• Non-contact behavioral studies are crucial in animal neuroscience and ethology.
• Deep learning advances enable intelligent image analysis for pose estimation and tracking.
• Investigating animal behavior in space presents unique challenges due to microgravity, radiation, and hypomagnetic fields.

Purpose of the Study:

• To address the lack of annotated image data for space-bound animals.
• To introduce the first multi-task, expert-validated dataset for analyzing multi-animal behavior in complex scenarios.
• To establish benchmarks for evaluating deep learning models in space animal behavior research.

Main Methods:

• Development of the SpaceAnimal Dataset, featuring model organisms like Caenorhabditis elegans, Drosophila, and zebrafish.
• Inclusion of expert-validated ground truth annotations for multi-animal behavior analysis.
• Provision of evaluation code for deep learning models.

Main Results:

• The SpaceAnimal Dataset is the first of its kind for multi-animal behavior analysis in space-related conditions.
• Established benchmarks for deep learning model performance in this domain.
• Facilitated the development of AI tools for analyzing complex animal behaviors.

Conclusions:

• The SpaceAnimal Dataset is a significant resource for advancing AI in space animal behavior research.
• This work will accelerate the discovery of new behavioral patterns in animals under space conditions.
• The dataset and evaluation code will foster innovation in artificial intelligence for biological research.