LabelMe

LabelMe is a web-based image annotation tool and dataset used in computer vision research. It enables users to outline objects with polygonal borders and assign category labels, producing structured annotations that describe object boundaries and classes. The resulting data can be exported in a JSON format, facilitating integration with machine learning workflows for tasks such as segmentation and recognition.

The project originated at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) in the mid- to late-2000s. It provided an online interface for uploading images, annotating objects, and sharing results, with the goal of crowdsourcing labeled data. The accompanying dataset aggregates images and polygonal annotations contributed by users, making it a widely used resource for vision research.

LabelMe has influenced subsequent annotation tools and datasets by popularizing polygon-based segmentation and a JSON-based annotation structure. Its open approach to data collection and annotation formats has made it a reference point in discussions of image labeling workflows. The platform and dataset have seen continued use in academic settings and have inspired variations and extensions in image annotation practices, education, and tooling.

Overall, LabelMe remains notable for providing accessible tools and data that support research in object detection, segmentation, and scene understanding, emphasizing openness and reusability for non-commercial educational and scholarly activities.