Thanks to the excellent work of NeuRAD, we reproduce some results on the Waymo open dataset.
Our goal in reproducing and open-sourcing this waymo dataparser for NeuRAD is to provide a basic reference for the self-driving community and to inspire more work.
In the same folder, there is wod_dataparser.py which followed the README-Adding Datasets suggestions. In addition, we added also wod_utils.py which did the main work for converting/exporting Waymo dataset.
In addition, we have also added the rolling shutter support for Waymo dataset as the rolling shutter direction is horizontal instead of the vertical one in Pandaset. Here are some examples of the comparison results (on squence of 10588):
Dataset | Sequence | Frames | Cameras | PSNR | SSIM | LIPS |
---|---|---|---|---|---|---|
Pandaset | 006 | 80 | FC | 25.1562 | 0.8044 | 0.1575 |
Pandaset | 011 | 80 | 360 | 26.3919 | 0.8057 | 0.2029 |
Waymo | 10588771936253546636 | 50 | FC | 27.5555 | 0.8547 | 0.121 |
Waymo | 473735159277431842 | 150 | FC | 29.1758 | 0.8717 | 0.1592 |
Waymo | 4468278022208380281 | ALL | FC | 30.5247 | 0.8787 | 0.1701 |
Notes: All above results were obtained with the same hyperparameters and configurations from NeuRAD paper (Appendix A)
Up is ground truth, bottom is rendered.
Left is ground truth, right is rendered.
Left is ground truth, right is rendered.
Left is ground truth, right is rendered.
Results has been done with waymo open dataset v2.0.0, gcloud link
- Lei Lei, Leddartech
- Julien Stanguennec, Leddartech
- Pierre Merriaux, Leddartech