MyVision is a free online image annotation tool for generating computer vision training data, developed as an open-source project on GitHub. It runs in the browser and focuses on making large-scale labelling fast and organised, with export to the dataset formats used by popular detection frameworks. The repository has around 610 stars and remains under active development in 2026.
Training an object detector starts with labelled images, and hand-drawing boxes for thousands of photos is slow, tedious work. MyVision streamlines that pipeline: you load images, draw bounding boxes or polygons with editing tools for adding, moving and removing points, assign labels, and export the result in formats such as COCO, YOLO, Pascal VOC and VGG. To cut the manual effort further, it can auto-annotate images with the COCO-SSD model โ and because the model runs locally in your browser via TensorFlow.js, your images never leave your machine, which keeps private datasets private. The hosted version at myvision.ai needs no installation.
Projects stay organised as they grow, and labels remain consistent across images in a dataset, which matters when several people label the same collection. For teams validating an idea, the combination of manual annotation, model-assisted pre-labelling and one-click exports means a usable detection dataset can be assembled in an afternoon and fed straight into a YOLO or TensorFlow training run.
MyVision is free to use online, and the source code is open under the GPL-3.0 license, so you can also host it yourself at no cost.
ML practitioners, students and small teams who need labelled image datasets for object detection and want a quick, privacy-friendly annotation tool without enterprise pricing.