Towards Robust Automated Mouse Scratching Behavior Detection for Itch Research
Automated detection of mouse scratching behavior could reduce manual annotation and is essential for preclinical itch research. However, robust performance across laboratories remains a challenge for current methods. We present Scratch-AID+, a framework dedicated to automated mouse scratching detection, including a diverse and progressive set of approaches: classic ResNet-based detectors,…
Automated mouse scratching behavior detection holds promise for reducing manual annotation in preclinical itch research. However, current methods struggle to maintain robust performance across laboratories. The research team introduces Scratch-AID+, a framework that combines classic ResNet-based detectors, task-adaptive fine-tuning of video foundation models, and in-context learning with multimodal large language models.
To assess real-world applicability, the researchers benchmark on a public dataset and test additional cross-laboratory video recordings. They discover that classic baselines struggle with performance degradation on external data, especially with sparse scratching events. Task-adaptive fine-tuned video foundation models prove effective, improving detection significantly.
Moreover, frozen multimodal large language models show a promising out-of-the-box capability, requiring only few-shot demonstrations. The team also explores the impact of diverse video preprocessing strategies. Their findings emphasize the enhanced performance and generalizability of adapted foundation models for automated scratching behavior analysis, paving the way for their real-world deployment.
The code for Scratch-AID+ is made publicly available at https://anonymous.4open.science/r/scratch aid-plus/.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
