Unified Dataset for the paper "Surgical instrument tracking: a lightweight YOLO approach across diverse domains"

This dataset provides a unified collection of surgical instrument images and YOLO-format annotations (bounding boxes), developed for the paper "Surgical instrument tracking: a lightweight YOLO approach across diverse domains" accepted at ICCSA 2026. The dataset was created by standardizing and merging three heterogeneous public surgical datasets: CholecTrack20 (Laparoscopic cholecystectomy) ROBUST-MIS (Colorectal surgery) Badilla-Solórzano (Synthetic/Dental context) It includes images and corresponding labels (.txt) formatted for training YOLO architectures to detect and track surgical tools across diverse clinical domains, addressing class imbalances and varied lighting conditions. Update (Version 1.1.0): Added YOLO_RESULTS.zip containing supplementary evaluation results, PR curves, and qualitative tracking figures that exceed GitHub's repository size limits. Update (Version 1.2.0): Added zero-shot generalization evaluation materials. This includes inference results on unseen surgical domains and clinical settings to demonstrate the model's robustness and cross-domain interoperability, as discussed in the ICCSA 2026 paper.