Thank for your great work!!!, but I have several Question, hope to get reply!
In Table 1, do these four lines use depth as supervision?
The configuration files all use depth as supervision.
But if lidar point cloud is used as depth supervision, will the results of BEVDet* and BEVDet4D not match the official results of BEVDet?
For the line BEVDet4D*, 4D and CBGS are used, and the experimental results should be as shown below.
However, the results in Table 1 are much lower than this, and it seems more like the result of not using lidar point cloud as supervision.
And I ran an experiment with the configuration file tig_bev4d-r50-CBGS.py, and the results are as follows:
Thank for your great work!!!, but I have several Question, hope to get reply!
In Table 1, do these four lines use depth as supervision?
The configuration files all use depth as supervision.
But if lidar point cloud is used as depth supervision, will the results of BEVDet* and BEVDet4D not match the official results of BEVDet?
For the line BEVDet4D*, 4D and CBGS are used, and the experimental results should be as shown below.
However, the results in Table 1 are much lower than this, and it seems more like the result of not using lidar point cloud as supervision.
And I ran an experiment with the configuration file tig_bev4d-r50-CBGS.py, and the results are as follows:
