about waymo result

[mc171819]

Hi, you mentioned you use adabin trained on waymo. So how you do that, since waymo don't provide the gt depth map. Another question is did you train the total model, depthcompletion, 2d detection and 3d detection in an end-to-end manner?

[Willy0919]

We trained the AdaBins[https://github.com/shariqfarooq123/AdaBins] using sparse depth maps generated by point clouds as ground truth. Since we focus on the 3d detection in our paper, the depth estimation and 2d detection are completed offline.

[mc171819]

We trained the AdaBins[https://github.com/shariqfarooq123/AdaBins] using sparse depth maps generated by point clouds as ground truth. Since we focus on the 3d detection in our paper, the depth estimation and 2d detection are completed offline.

thank you, so you use sparse depth map as a dense one to train adabin, using the same loss as adabin?