This paper was originally presented at the 4th International Workshop on Camera Traps, AI, and Ecology, 5th - 6th September 2024 in Hagenberg, Austria and Online Here you can find the videos and annotations of the presented AUTH-Sheep dataset in COCO format. Each video consists of: - mp4-video in 3840x2160 @30 fps - "instances_default.json" - including all annotations for all classes, directly downloaded from CVAT - "annotations_hbb.json" - annotated sheep with axis aligned boxes, loadable with pycocotools - "annotations_obb.json" - annotated sheep with oriented boxes, loadable with pycocotools (segmentation) New oriented annotatons for SheepCounter and useful scripts can be found at https://github.com/idmt-odoll/AUTH-Sheep