Bl-Vision provides tools for generating synthetic image datasets directly within Blender. The addon creates 2D bounding boxes for objects, collections, and particle systems, exporting the labels in standard YOLO or COCO formats. It includes options for raycasting to automatically ignore occluded items from the camera view.

The tool outputs paired rendered images alongside annotation files, streamlining the creation of training data for computer vision applications. Users can assign category IDs manually or use an automated assignment function before rendering the animation sequence to generate the dataset.