Cattle Behavior Recognition Dataset

The current agricultural industry faces challenges in monitoring and managing cattle behavior, with traditional methods relying on manual observation, which is inefficient and error-prone. Existing automated monitoring systems still have shortcomings in accuracy and real-time performance, unable to meet the needs of modern smart farming. This dataset aims to provide high-quality cattle behavior recognition data to support the training of target detection models and improve the accuracy of behavior recognition. The dataset contains 15,000 annotated images, collected using high-resolution cameras in natural breeding environments to ensure image quality. We use multiple rounds of annotation and expert review for quality control to ensure data consistency and accuracy. All images are stored in JPG format, organized by image ID for easy subsequent processing and analysis. The dataset has a clear structure, including key fields such as image ID, bounding box, and behavior labels, meeting the needs of target detection tasks.