The authors say it combines behavioral labeling with individual localization and produces structured outputs for training and validating computer-vision models. The paper describes validation using 56 hours of wild lemur footage. Its focus is interaction annotation, rather than only detecting individual actions, and the authors note possible use in human-interaction video analysis. The supplied source does not establish performance across other datasets or settings.
Validated on 56 hours of wild lemur footage with ten interaction classes; paper here .
In one minute
SiLVi is an open-source video-labeling tool for annotating animal behavior and interactions.
What changed: The paper reports SiLVi, an open-source video annotation tool for labeling animal behaviors and interactions; no prior document covers this work.