Tai-Wang

So it might be a bug related to the inference process? We will first try to reproduce your problem soon. BTW, what is the commid id of mmdet3d are you...

We will fix this issue in #1457 .

Please refer to this [tutorial](https://mmdetection3d.readthedocs.io/en/latest/tutorials/pure_point_cloud_dataset.html) for the way of converting pcd file to bin.

You may add breakpoints in your codes and have a look at if there is any boxes output. Maybe your model is not trained well, or there is some inconsistency...

It should not be related to the refactoring of coordinate systems because PointRCNN is merged after the release of 1.0.0rc. "couldn't get the reported performance" with training from scratch means...

Please "resolve" the previous comments when you fix them, and then we can have the next round review. Thanks for your great efforts!

Please fix the linting. You can refer to the anno_file format of KITTI. It saves the filename, calibration and pose information in the anno_file such that we can directly load...

Please fix the linting. You can refer to the anno_file format of KITTI. It saves the filename, calibration and pose information in the anno_file such that we can directly load...

@ZCMax Please have a look at this issue and update the log files if necessary.

Thanks for your report. @ZCMax Please reproduce this problem and fix it if necessary.