xby-obb

Standard object detection models output axis-aligned boxes, which fall short when the targets are inherently rotated — planes, ships, vehicles, sports courts — producing loose boxes and losing the very orientation information that downstream tasks depend on. xby-obb bridges this gap by integrating the YOLO26 oriented bounding box (OBB) model directly into DeepSeek Harness, so a single chat call returns the rotated box, heading angle, confidence score, and class label without any separate inference pipeline to maintain.\n\nThe plugin accepts images through three input modes — a file URL, a base64-encoded string, or a local file path — covering online data, in-memory transfer, and offline batch workflows alike. Detection covers 15 target classes including plane, ship, storage tank, baseball diamond, tennis court, basketball court, ground track field, harbor, bridge, large vehicle, small vehicle, helicopter, roundabout, soccer ball field, and swimming pool, which together span the most common oriented objects in remote-sensing imagery and open-scene surveillance.\n\nIf your work involves remote-sensing interpretation, automatic sports-venue annotation, vehicle-orientation analysis in security feeds, or any vision task that requires precise rotated bounding boxes, xby-obb lets you skip the model-deployment overhead and go straight from image to structured results inside the conversation.