Get the latest tech news

Apple's Cubify Anything: Scaling Indoor 3D Object Detection

Contribute to apple/ml-cubifyanything development by creating an account on GitHub.

For each frame in each capture, we include "per-frame" 3D box ground-truth which was produced using the rendering process outlined in the Cubify Anything paper. An additional file is included as[video_id]/world.gt/instances.json which corresponds to the full world set of 3D annotations from which the per-frame labels are generated from. We support visualizing the predictions on the GT point cloud (derived from Faro depth) when using the--viz-on-gt-points flag.

Get the Android app