CARLA-4D Visualizations for Revealing Occlusions with 4D Neural Fields

Main Page More Visualizations GREATER Download Dataset

Description

Our model takes a point cloud video clip with 12 frames as input, and subsequently conditions a spatiotemporal neural field in order to predict output point clouds of the complete dynamic scene at a chosen moment in time. Note that these results are non-cherry-picked as they comprise the first 20 scenes of the test set.

Geometry Completion + Semantic Completion

Test Scene 1
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 2
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 3
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 4
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 5
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 6
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 7
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 8
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 9
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 10
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 11
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 12
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 13
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 14
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 15
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 16
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 17
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 18
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 19
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud
Test Scene 20
Input RGB Video
Input Point Cloud
Output Semantic Point Cloud
Target Semantic Point Cloud