Helix4D's Inference on a random subset of HelixNet

3D point cloud with colors corresponding to :

  • inference with the default configuration of Helix4D
  • ground truth
  • slices used for inference
  • partition of the point cloud at the coarse level
  • point acquisition time
The model is trained with full rotations of the sensor, and inference is done on 1/5 turn.
Warning: 3D scenes might take some time to load.
For visualization, all point clouds are downsampled on a regular grid of size 25cm.
You can left click to rotate, right click to translate, scroll to zoom, and change the visualized layer with the menu.