10.7K ·GitHub
What is this?
A free cross-platform desktop app for AI-assisted annotation of text, image, video, and multimodal data — manual labeling tools plus built-in deep learning models that pre-label or batch-annotate, with multi-format export.
A free cross-platform desktop app for AI-assisted annotation of text, image, video, and multimodal data — manual labeling tools plus built-in deep learning models that pre-label or batch-annotate, with multi-format export.Is it a desktop app or does it run in the browser?It is a native desktop application, not a web service — it installs on Windows, Linux, and macOS, with a GUI available in English, Simplified Chinese, Japanese, and Korean.
It is a native desktop application, not a web service — it installs on Windows, Linux, and macOS, with a GUI available in English, Simplified Chinese, Japanese, and Korean.How do I get it running?Follow the Installation & Quickstart guide linked in the README's Docs section — you install a desktop build for your OS. A command-line interface is documented as well for scripted workflows.
Follow the Installation & Quickstart guide linked in the README's Docs section — you install a desktop build for your OS. A command-line interface is documented as well for scripted workflows.Do I need a powerful GPU or paid API keys for the AI-assisted labeling?Core models run locally through ONNX Runtime, TensorRT, or OpenCV DNN — no API keys required. Optionally, you can connect cloud vision-language models (Gemini, ChatGPT, GLM) or route work to a remote X-AnyLabeling-Server.
Core models run locally through ONNX Runtime, TensorRT, or OpenCV DNN — no API keys required. Optionally, you can connect cloud vision-language models (Gemini, ChatGPT, GLM) or route work to a remote X-AnyLabeling-Server.Does my dataset stay on my machine, or is it uploaded somewhere?By default it stays local — inference runs on your own hardware. Data only leaves the machine if you opt into remote inference via X-AnyLabeling-Server or cloud model backends.
By default it stays local — inference runs on your own hardware. Data only leaves the machine if you opt into remote inference via X-AnyLabeling-Server or cloud model backends.Is the license OK for commercial annotation work?It is GPLv3 — you may use, modify, and redistribute the app, including for commercial purposes, as long as you follow the license terms.
It is GPLv3 — you may use, modify, and redistribute the app, including for commercial purposes, as long as you follow the license terms.What annotation tasks does it cover?A wide range: image classification, object detection with regular and rotated boxes, instance and semantic segmentation, pose and face estimation, multi-object tracking, OCR and key information extraction, lane annotation, depth estimation, image matting, tagging, captioning, visual question answering, and document parsing. Drawing tools include polygons, rectangles, rotated boxes, cuboids, circles, lines, polylines, points, and masks.
A wide range: image classification, object detection with regular and rotated boxes, instance and semantic segmentation, pose and face estimation, multi-object tracking, OCR and key information extraction, lane annotation, depth estimation, image matting, tagging, captioning, visual question answering, and document parsing. Drawing tools include polygons, rectangles, rotated boxes, cuboids, circles, lines, polylines, points, and masks.Which AI models does it ship with?Dozens across tasks: the YOLO family from v5 through YOLO26 for detection, segmentation, and pose; DETR-style models like RT-DETR, RF-DETR, and D-FINE; SAM 1/2/3 plus lightweight variants such as MobileSAM and EfficientViT-SAM; Grounding DINO and YOLO-World for text-prompted detection; PP-OCRv4/v5/v6 and PaddleOCR-VL for OCR and document parsing; Depth Anything; and vision-language models including Qwen3-VL, Florence2, Gemini, ChatGPT, and GLM.
Dozens across tasks: the YOLO family from v5 through YOLO26 for detection, segmentation, and pose; DETR-style models like RT-DETR, RF-DETR, and D-FINE; SAM 1/2/3 plus lightweight variants such as MobileSAM and EfficientViT-SAM; Grounding DINO and YOLO-World for text-prompted detection; PP-OCRv4/v5/v6 and PaddleOCR-VL for OCR and document parsing; Depth Anything; and vision-language models including Qwen3-VL, Florence2, Gemini, ChatGPT, and GLM.What dataset formats can I import and export?It reads and writes COCO, VOC, YOLO, DOTA, MOT, MASK, PPOCR, MMGD, VLM-R1, and ShareGPT, so annotations move straight into common training pipelines.
It reads and writes COCO, VOC, YOLO, DOTA, MOT, MASK, PPOCR, MMGD, VLM-R1, and ShareGPT, so annotations move straight into common training pipelines.Can I plug in my own model?Yes — the app supports custom model integration, flexible extension, and secondary development, with a dedicated 'Customize a model' guide. For serving your own backends, the companion X-AnyLabeling-Server connects custom models and compute resources for remote inference.
Yes — the app supports custom model integration, flexible extension, and secondary development, with a dedicated 'Customize a model' guide. For serving your own backends, the companion X-AnyLabeling-Server connects custom models and compute resources for remote inference.What are similar projects?ultralytics, deep-learning-for-image-processing, SCM, yolov5, supervision, video-subtitle-remover, detectron2, March7thAssistant
ultralytics, deep-learning-for-image-processing, SCM, yolov5, supervision, video-subtitle-remover, detectron2, March7thAssistantWhich topics is it in?paddlepaddle, onnxruntime, instance-segmentation, annotation, opencv, image-segmentation, yolo, pose-estimation, face-detection, segmentation, vlm, vision-language-model
paddlepaddle, onnxruntime, instance-segmentation, annotation, opencv, image-segmentation, yolo, pose-estimation, face-detection, segmentation, vlm, vision-language-model