Kuan-Chih Huang

> > Hi, I downloaded nuscenes dataset and when I run the convert_nuscenes.py, it told me a file doesn't exist in the dataset (from what I remember this file is...

@qiank10 Thanks for your reply! I have some question: 1. What's the foggy beta in the table 1 2. How can I run the MVDNet-fast model Thank you~ Best regards.

Paper name/title: PTT: Point-Trajectory Transformer for Efficient Temporal 3D Object Detection Paper link: [https://arxiv.org/abs/2312.08371](https://arxiv.org/abs/2312.08371) Code link: [https://github.com/kuanchihhuang/PTT](https://github.com/kuanchihhuang/PTT)

Hi @AlfredQin, 1. The released checkpoint was trained on our refactored code. 2. The configuration is similar. We have tested on our local environments (V100 with torch 1.4 / 3090...

Hi @AlfredQin, The released model was trained on the current codebase. There is no setting/config difference after refactoring, just for better readability. Maybe you can change batch size or epoch...

Hi, the depth ground truth looks reasonable. Can you comment out the code for depth loss to check the training for image-only works well or not?

Hi, depth loss looks to work well. The problem seems to come from roty and alpha (according to your observation). Can you try to plot the orientation loss? Also, most...