AI Models Developed for Animal Detection and Pose Estimation, Including Cows
A new GitHub repository, 'ComfyUI-WanAnimalPreprocess,' has introduced AI-powered models designed for animal detection and pose estimation. These models utilize YOLOv8 ONNX for detecting various animal classes, including cows, and ViTPose models for 17-keypoint animal pose estimation. The system is built as custom nodes for ComfyUI, adapting human pose estimation technology for animal use. Users can select between AP10k and APT36k datasets, with `yolov8m.onnx` or `yolov8l.onnx` recommended for optimal speed and accuracy in detection. The ViTPose models come in various sizes, offering different balances of speed and quality, from `vitpose-s-apt36k.onnx` for fastest processing to `vitpose-h-apt36k.onnx` for best quality. The system supports a range of animals beyond cows, such as cats, dogs, horses, sheep, elephants, bears, zebras, giraffes, and birds.