Tai-Wang

So do your experiments show that the LiDAR axis definition in lyft data is "x-back, y-right, z-up"? @bmankirLinker

Maybe some functions are not affected by its axis definition, but we are indeed encountered bugs when inferencing orientations of objects by using some models adapted from NuScenes/KITTI. BTW, @joinssmith...

All right, hope an update...Thanks all the same~

Hi @PhanThanhTrung , thanks for your interest and feedback. Sorry that this code release is still a draft version and there are many details to be fixed. I do not...

Hi @PhanThanhTrung , I have fixed the monocular 3D detection demo to make it compatible with DfM. Please try it following [this instruction](https://github.com/Tai-Wang/Depth-from-Motion#demo) :)

我这两天再做下最后走通整个流程的验证，会在这周末前更新一版 code，请到时关注和试用~

之前尝试过可以缩小 [num_bins](https://github.com/Tai-Wang/Depth-from-Motion/blob/main/configs/dfm/dfm_r34_1x8_kitti-3d-3class.py#L6) 到 96 可以少掉一些点，以及可以尝试去掉[这个地方](https://github.com/Tai-Wang/Depth-from-Motion/blob/main/configs/dfm/dfm_r34_1x8_kitti-3d-3class.py#L79)的 2d detection head；再进一步缩小的话可能需要把 voxelization 放到 CPU 上做会好些。总体上之前有缩到 12G 显存内，重构后还没时间尝试，过段时间我可以看看，在那之前你可以先尝试这几点试试看。

KITTI 部分因为代码进行了重构，近期还在抽时间对训练精度，目前还稍微有些不稳定，待完全对齐会更新当前 readme，以及 release 全部 config/models。 Waymo 部分或许可以先尝试着跑一下，有什么问题可以反馈，应该 PGD 和 MV-FCOS3D++ 的 config 都是能够跑通的。不过数据预处理部分还没有仔细整理和验证过，我尽量在月底之前完成验证。

Hi @le-cheng，近期对项目内容进行了初步更新，可以尝试用最新的代码和配置文件跑下试试，如有问题欢迎再来提相应的 issue。

Hi @AndyYuan96 , thanks for your contribution! Please fix the lint failure (by using pre-commit hook following this [guide](https://github.com/open-mmlab/mmcv/blob/master/CONTRIBUTING.md)) and rebase your branch onto the `dev` instead of `master`. We...