rectlabel.com

What it is

RectLabel is an offline macOS image annotation tool developed by Japanese indie developer Ryo Kawamura, mainly aimed at creating datasets for object detection and segmentation. It is not a cloud-based crowdsourcing annotation platform; it is closer to a local, developer-oriented visual data preparation tool, suitable for organizing images or video frames into trainable datasets.

Core capabilities

Based on the available information, RectLabel supports a wide range of annotation types, including bounding boxes, polygons, pixels, rotated boxes, keypoint skeletons, Bézier curves, line segments, and points. For AI-assisted annotation, it integrates Segment Anything Model 2/3, supports annotating multiple objects at once via text and box prompts, and also supports Cellpose, tracking models, and Core ML auto-labeling capabilities including RF-DETR and YOLO26. It also provides automatic text recognition for line- and word-level text annotation. For export, it supports YOLO, COCO, CreateML, DOTA, as well as color-index masks and grayscale masks, making it easy to fit into common training workflows.

Pricing and deployment

The standard RectLabel version uses a subscription model, priced at $2.99/month or $9.99/year, with an unlimited-time free trial, though continued use requires purchasing a license. RectLabel Pro is a one-time purchase at $19.99, making it more suitable for long-term use and batch deployment. The Pro version also supports volume purchases through Apple Business Manager, and notes that it has no network connection because there are no in-app purchases. Both apps provide the same features.

Pros and cons

Its advantages are offline operation, low pricing, rich annotation types, and support for multiple mainstream training formats, making it friendly to individual developers, researchers, and small CV teams. Its AI-assisted annotation features can also reduce the cost of fully manual box drawing and segmentation. Limitations include no mention of Windows, Web, or multi-user collaboration versions, nor enterprise annotation platform features such as APIs, team quality control, task assignment, or permission management. Chinese UI and Chinese documentation support are also not clearly stated.

Who it is for and access from China

RectLabel is better suited to macOS users, computer vision engineers, researchers, and teams that need to quickly create YOLO/COCO/CreateML/DOTA datasets. The source text does not provide information about access from China, and purchases depend on the App Store or Apple Business Manager, so network and payment experience may depend on the user’s regional Apple account environment. If you need Web-based collaboration or open-source self-hosting, compare it with CVAT or Label Studio; if you need cloud training and data management, alternatives such as Roboflow and Labelbox may be worth considering.