LVT aims to reduce the time AV researchers spend dealing with dataset API issues so they can spend more time analyzing data. Researchers are able to load in data and immediately visualize it as LiDAR point clouds or RGB images.
Our app aims to provide researchers with the necessary tools to easily visualize the data their models are producing so that they can quickly find and see faults with their own eyes to get a better idea of why their model is failing.
-
Interactive GUI based scene navigation
-
Real time sensor switching
-
Automated fault scanning (False Positives, Incorrect Annotations, Unmatched GT Annotations)
-
Tested for Windows and Linux
-
Real time annotation filtering by class, confidence score, or error type
-
Export rendered figures to PNG
-
Modular system to support expansion to multiple dataset formats
