metadriverse / scenarionet

ScenarioNet: Scalable Traffic Scenario Management System for Autonomous Driving

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Errors when converting nuScenes dataset

jongchyisu opened this issue · comments

Hello, thanks for the great work!
I'm trying to convert the nuScenes dataset, however I got several errors:

  1. There are scenes with velocity >50 and causing the assertion error. Is there something wrong for velocity measurement or is the velocity correct?
  2. I then got error of ValueError: could not broadcast input array from shape (193,3) into shape (191,3) at this line https://github.com/metadriverse/scenarionet/blob/e9c1419a91c42e67fe176a538d2984cf1d9e0f93/scenarionet/converter/nuscenes/utils.py#L206C13-L206C81 What's the issue of this?
  3. How do you deal with multiprocessing's OOM error (shows "killed")?

Thanks for your time!

Hi, sorry for the late reply. I didn't watch this repo until now...

For your questions:

  1. nuScenes data is actually collected in a low frequency, so I interpolate the data to make them 10hz. To make sure the interpolation won't generate some outliers, I add that some assert. Just ignore them for free.

  2. Could you tell me which nuScenes split you are using? There must be something wrong in the interpolation process. It is currently tested on the nuPlan-mini only. I will fix the bugs and update the progress soon.

  3. Lower the number of workers can solve this like 'num_workers=2'