AI system detects fish loss of equilibrium from temperature stress

Researchers have unveiled an AI‑driven system that automatically and objectively identifies the precise moment fish lose equilibrium in response to temperature stress. By integrating DeepLabCut, a deep‑learning tool that tracks animal posture from video, with ResNet34, an image‑classification network, the new platform can pinpoint loss‑of‑equilibrium events in real time without human intervention.

The system was tested on several species of freshwater and marine fish exposed to controlled temperature ramps. DeepLabCut extracted fine‑grained postural data, while ResNet34 classified the resulting frames to detect the critical threshold where equilibrium is lost. The combined approach yields high accuracy and reproducibility, enabling large‑scale monitoring of fish behavior under varying thermal conditions.

Experts say the technology will be valuable for assessing how rising ocean and river temperatures affect fish populations. By providing objective, high‑throughput data on temperature‑induced behavioral changes, the tool can inform climate‑impact models and help fisheries managers develop adaptive strategies for vulnerable species.