Anylabeling

AnyLabeling is an open source desktop application for annotating images to build computer vision datasets, with AI assisted labeling from Segment Anything and YOLO models. Its maintainers describe it as LabelImg plus Labelme with an improved interface and automatic labeling, so a user can click a point or draw a box and let a segmentation model produce the mask instead of tracing it by hand. It suits machine learning engineers, researchers and data teams preparing training data for detection and segmentation models. Manual annotation covers polygons, rectangles, circles, lines and points. Automatic labeling supports YOLOv8 object detection and the Segment Anything family, including the original SAM models, MobileSAM, SAM 2, SAM 2.1 and SAM 3, which accepts text prompts for open vocabulary segmentation. The tool also handles text detection, recognition and key information extraction labeling. Model weights download automatically on first use, and hardware acceleration is available through CUDA, CoreML, DirectML, OpenVINO and vendor NPU providers. The interface is available in English, Vietnamese and Chinese. AnyLabeling is free. It installs from PyPI with pip on Python 3.11 or later, or as a prebuilt executable for Linux, Windows and Apple Silicon macOS, with separate CPU and GPU accelerated builds for each platform. The source is on GitHub. The same developer has released AnyLearning as open source, a companion project for labeling data and training models entirely offline, which the AnyLabeling site now promotes alongside the labeling tool.