visulisation of augmented data in multi-modality training

[cattpku]

Hi All,
I am trying to following instructions in https://mmdetection3d.readthedocs.io/en/latest/user_guides/visualization.html for a multi-modality dataset visualisation, expecially after augmentation.
The dataset used is KITTI, and command I used is (exactly the same)
python tools/misc/browse_dataset.py configs/mvxnet/mvxnet_fpn_dv_second_secfpn_8xb2-80e_kitti-3d-3class.py --task multi-modality_det --output-dir ${OUTPUT_DIR}
but I found the output is different from the images in the doc, if augmentation is not applied, groundtruths for both RGB and Lidar are correctly plotted, while once augmentation is applied, groundtruth for Lidar is correctly plotted but groundtruth for RGB is wrong.
Here is one incorrect example:


It is really appreciated if anyone could advise what's going on and how to fix it. Thanks.