Guangda Ji comments

Results 8 comments of


                                            Guangda Ji

Config to train on multiple datasets

新年快乐! 万事如意!

Config to train on multiple datasets

Hi, I would like to ask if there is any updates on the joint training config. Thanks!

Extend to wild videos?

The current pipeline needs camera's pose, intrinsic, depth and RGB.

Missing id to class mapping

I think this function will give you the correct id [here](https://github.com/cvg/LabelMaker/blob/ba56b548156d76d899fdd715fdcb0d4a50f03058/labelmaker/label_data.py#L1395)

Camera Convention

Check [here](https://github.com/cvg/LabelMaker/blob/6dd6d72059d42efc7bab93d2526083753a8d6daa/scripts/utils_3d.py#L85) it is the inverse of extrinsic

Can't find python inside the docker container

I think you have to set the path of conda (I can't remember it well, maybe /miniconda3/bin/conda) into environment path. An example wit here: https://github.com/cvg/LabelMaker/blob/b3397d0be8897c8fcf7bf83b46c26e7d5f9bfe9e/pipeline/activate_labelmaker.sh#L1

Results change during testing / evaluation

I remember in the pre-processing code of pointcept, they use some sort of sampling as data augmentation. Therefore, the evaluation score during the training is noisy. Please check in detail...

Results change during testing / evaluation

I am sorry I cannot help you with this as the code of PTv3 and PPT are contributed from the original authors.