remyxai / VQASynth

Compose multimodal datasets 🎹
https://twitter.com/smellslikeml/status/1756723056675094726
216 stars 13 forks source link

Adjust point cloud viz + localizer fallback #25

Closed salma-remyx closed 2 weeks ago

salma-remyx commented 2 weeks ago

To help create better point cloud visualizations, we update the SpatialSceneConstructor's run() method to colorize segmented point clouds.

We also add a resize image fallback strategy to avoid tensor mismatches running with CUDA when inputs/outputs are too large.