Haiyang Wang

If needed, everyone may try the [GiT repository](https://github.com/Haiyang-W/GiT), a general end-to-end vision transformer, which fully covers the tasks included in visionLLM and can also handle semantic segmentation. The code and...

Thanks very much! I'll check this part and fix it as soon as possible. :) Thanks a lot for your contribution. :)

Sorry for the late reply. Please refer to [here](https://github.com/Haiyang-W/CAGroup3D?tab=readme-ov-file#data-preparation) for generating your own dataset. Note that the coordinate system adopted in MMdetection3D(0.15) is very different from MMdetection3D(>=1.0).

Our data preprocessing strictly follows the mmdetection3d (0.15) version. The inconsistency should only be in the coordinate system compared to the existing mmdetection3d.

Any update? I also meet this problem.

I think it should only require aligning the dataset config, organizing aspects like dataset range and orientation angle. It may need a closer look.

In our paper, we did not conduct experiments on joint training for the two tasks; they were trained separately. However, both configs are provided, so it should be relatively easy...

Thank you for your contribution. I am currently busy with my PhD defense, which may take another week or two. After that, I will do some code review if I...

UniTR starts fusion at the backbone level, so there is no separate concept of LiDAR features and image features. However, if you want to obtain the LiDAR tokens and image...

> Hi, [@Haiyang-W](https://github.com/Haiyang-W) Thanks for your work. I noticed the lidar-only and camera-only results in your paper, and I'd like to learn about the cfg files and models. Thanks a...