// the find
CVHub520/X-AnyLabeling
Effortless data labeling with AI support from Segment Anything and other awesome models.
X-AnyLabeling is a desktop annotation tool built on PyQt6 that integrates a large collection of CV models (SAM, YOLO variants, Grounding DINO, OCR, depth, tracking) to speed up labeling workflows. It targets ML practitioners who need to build training datasets and want model-assisted annotation rather than purely manual drawing. Actively maintained by a single developer who keeps pace with new model releases.
The model library breadth is genuinely impressive — SAM 1/2/3, every YOLO generation through v26, Grounding DINO, Florence2, Qwen3-VL, and more, all runnable locally via ONNX or TensorRT without a cloud dependency. Export format coverage (COCO, VOC, YOLO, DOTA, MOT, ShareGPT, VLM-R1) means you're not locked into one training pipeline. The remote inference server companion project is the right architectural move for teams sharing GPU resources. The annotation check status workflow added in April 2026 is a practical quality-of-life feature that annotation tools almost universally skip.
Single-maintainer bus factor is a real risk — the donation-driven sustainability model and WeChat contact for enterprise questions don't inspire confidence for production dataset pipelines. The ONNX-first approach means you're often running quantized or community-converted weights, not the official ones, and model quality can silently degrade. No team/multi-user workflow: no label assignment, no review queue, no conflict resolution — this is a solo annotation tool that happens to support many models, not a labeling platform. The GPL-3.0 license (the README incorrectly calls it LGPL-v3 in the badge) will block commercial closed-source use for anyone who reads the actual LICENSE file.