Update training setting/preparation for KITTI

[phongphuhanam]

Hello,
Thank for releasing source code for your remarkable work.
BTW, I know it is still in progress, could you release/provide the training data setup for KITTI dataset as well.
I appreciate your consideration.

[ZijiaLewisLu]

Hi,

Thanks for your interest. I am occupied by a few other things and may need some time. I am wrapping up on personpath and will continue to kitti next.

For a quick explanation, we use the same detection as OC Sort. After you have the detection, the rest of the preprocessing and training should be similar as mot17 dataset.

[phongphuhanam]

@ZijiaLewisLu Thank you for answering my question.
I'm reading the code to re run the training step and validation step for MOT17.
As for KITTI, according to your suggestion, Do I need to convert KITTI data format to MOT17 data format and use OC Sort detection as the input.

[phongphuhanam]

@ZijiaLewisLu I'm able to re-run the training and validation steps for MOT17 and Person Path.
I realize there's quite a lot of prep involved in running the training and validation.
I know that you're busy at the moment, but could you provide the preparation scripts similar to those used for
MOT17 and Person Path?