Annotated Dataset for Orientated Surgical Instrument Detection in Hysterectomy Procedures

This dataset provides annotated images for oriented surgical instrument detection in hysterectomy procedures. The images were collected from both a controlledphantom surgical platform and real-world clinical operations. The dataset contains 4,404 RGB images with a total of 11,313 labeled instrument instances exhibiting varying sizes, orientations, and visual appearances. Among these, 3,975 images (10,218 instances) are used for training, while the remaining images are reserved for validation. All images are stored in the `images` folder, and the corresponding annotations are provided in the `labelTxt` folder. Each row in an annotation file corresponds to a single instrument shaft(-s) or tip(-t) instance. The annotation format is: [x1, y1, x2, y2, x3, y3, x4, y4, instrument_label] where (x1, y1)–(x4, y4) denote the coordinates of the four vertices defining the oriented bounding box of the instrument region, and `instrument_label` specifies the instrument category. This dataset is intended to support research in oriented surgical instrument detection, computer vision for minimally invasive surgery, and perception for surgical robotics.