This study introduces the first AI model for automated detection of acute postoperative pain in rabbits, addressing a gap in pain assessment for this widely used experimental model and popular pet. Utilizing a dataset of video footage from 28 rabbits before and after surgery, the researchers developed an AI model that leverages both facial expressions and body posture. A key methodological contribution is a two-step approach that incorporates temporal information from video data without heavy computational burden, employing 1-second interval sampling and Grayscale Short-Term stacking (GrayST). Furthermore, a novel frame selection technique was applied, which selects the most informative frames based on classifier confidence levels, effectively reducing ‘noise’ from low-quality or obstructed video data. The model achieved an accuracy of over 87%, with the best performance observed when using the CLIP/ViT + Naive Bayes backbone combined with GrayST and the top-frame selection method. This approach not only improves accuracy but also effectively handles challenges like occlusion and varied camera angles often present in real-world animal video data.
Non-Invasive Computer Vision-Based Fruit Fly Larvae Differentiation: Ceratitis capitata and Bactrocera zonata
This paper proposes a novel, non-invasive method using computer vision
