X-AnyLabeling

X-AnyLabeling wraps a full annotation studio around model-assisted labeling, covering horizontal and rotated boxes, polygons, instance and semantic masks, keypoints, OCR text, and multi-object tracking across images and video. Its auto-labeling engine loads SAM 1 through 3, SAM-HQ, and EdgeSAM for interactive segmentation, YOLOv5 through YOLO26 and YOLO-World for detection, Grounding DINO for open vocabulary prompts, and vision language models such as Qwen3-VL and Florence-2, with inference running through ONNX Runtime, OpenCV DNN, or TensorRT on NVIDIA GPUs. The desktop GUI runs on Windows, Linux, and macOS, installed from PyPI as x-anylabeling-cvhub on Python 3.11 or newer or launched from source, and exports flow to COCO, YOLO, VOC, DOTA, MOT, and other training formats. The project is GPL-3.0 licensed, which permits commercial use but requires derived tools to remain open. Actively developed by CVHub with commits into late July 2026 and roughly 9,900 GitHub stars, it is one of the most capable open alternatives to commercial labeling platforms.