Goh ELK

ContrastivePose: A contrastive learning approach for self-supervised feature engineering for pose estimation and behavorial classification of interacting animals

- Study of interaction behaviors in animals is a useful tool in neuroscience research. - Automated behavior analysis is enabled by machine learning using pose estimation. - Training supervised classification models faces bottleneck of manual labeling. - Contrastive learning improves model performance with unlabeled pose data.