biigle.de

What it is

BIIGLE is an online web-based annotation service for images and videos. It was originally built for marine environmental monitoring and exploration, but its documentation says it can also be used for general image and video annotation tasks. According to the official site, BIIGLE has 42 million annotations, 6,000 users, and 7.5 million images and videos. Its positioning leans more toward research and large-scale visual data management than a lightweight general-purpose annotation tool.

Core capabilities

For annotation, BIIGLE supports points, rectangles, circles, lines, and polygons. For video, it also supports whole-frame annotations and timeline-based object annotations across frames. For collaboration, users can invite colleagues into projects, create hierarchical label taxonomies, and edit, version, and share labels, with different permission levels available. The Largo label review grid is designed for quickly browsing large volumes of annotations, making it useful for quality control and training-data review. BIIGLE can also view and annotate very large gigapixel images directly in the browser.

AI features and limitations

BIIGLE’s main AI highlights are MAIA, Magic SAM, and smart sorting. MAIA uses novelty detection to identify candidate “interesting” objects. After human filtering, it trains a task-specific detection model and then generates candidate annotations for humans to confirm and label. In paper-based testing, it detected an average of 84% of objects of interest. Magic SAM uses the Segment Anything Model for fast segmentation. It is important to note that MAIA is not fully automatic: it can miss objects, produce false positives, and relies on human filtering. Its novelty detection is better suited to scenarios with relatively uniform backgrounds and sparse target objects. Tasks may run for several hours or even a full day, and parameter settings can affect the results.

Pricing, API, and privacy

Pricing is very friendly: the source code is free and open source under GPL-3.0, the public instance at biigle.de is free to use thanks to support from de.NBI Cloud, and self-hosting is also possible. The official site mentions a comprehensive API, which can be integrated into analysis pipelines and custom workflows. On privacy, user content remains owned by the user, but BIIGLE receives the access required to provide the service. The terms also explicitly state that, for technical reasons, BIIGLE may access files, annotations, and metadata being processed. Users handling sensitive data should therefore prioritize evaluating self-hosting.

Pros, cons, and who it is for

BIIGLE’s strengths are that it is free and open source, supports complex image and video annotation, offers practical AI assistance, and has well-developed collaboration and label-management features. Its drawbacks are that Chinese-language support is not disclosed, MAIA has a relatively high learning and configuration cost, and information on service guarantees and commercial support for the public instance is limited. It is a good fit for research institutions, marine monitoring teams, and data teams that need large-scale visual annotation and quality control. For use in China, there is no textual basis for assessing official-site access, account registration, upload speed, or cross-border data compliance, so its accessibility should be treated as unknown. Alternatives to compare include CVAT, Label Studio, Supervisely, and Roboflow Annotate.