Reproduce “Video-based estimation of pain indicators in dogs” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Hongyi Zhu, Yasemin Salgırlı, Pınar Can, Durmuş Atılgan, Albert Ali Salah

Researchers from Utrecht University and Ankara University developed a two-stream deep neural network to automatically estimate pain in dogs using video analysis, integrating both visual and kinematic features. The system achieved an F1-score of 76.3% on a newly created, veterinary-annotated dataset of dog behaviors.

Key insights

  • Integrating both spatial-temporal appearance features (from RGB video) and dynamic kinematic patterns (from normalized body keypoints) improves dog pain estimation.
  • Pre-trained pose estimation models, despite being trained on general animal datasets, can be effectively adapted for dog-specific pain analysis after robust preprocessing.
  • The creation of specialized, veterinary-annotated datasets is fundamental for advancing machine learning applications in animal welfare, addressing a significant data scarcity.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.