Datature
8 Case Studies
A Datature Case Study
Ultralytics, the developer behind the popular YOLO model series, needed to train a custom YOLOv8 object detection model to identify boats. Their challenge was preparing a properly annotated dataset in the specific format required for YOLO model training. They used Datature Nexus, an annotation platform, to efficiently create and export the necessary training data.
Datature provided the solution through its Nexus platform, which was used to upload, manage, and complete the annotation of 500 images with bounding boxes. The annotations were then exported in the precise YOLO Darknet format, including a 70/30 train-validation split. This prepared dataset allowed Ultralytics to successfully train their custom YOLOv8 model using their own Python SDK, enabling object detection on both images and video.