Question about Argoverse validation dataset order

tusen-ai / SST

Code for a series of work in LiDAR perception, including SST (CVPR 22), FSD (NeurIPS 22), FSD++ (TPAMI 23), FSDv2, and CTRL (ICCV 23, oral).

Apache License 2.0

765 stars 96 forks source link

Closed seonhoon1002 closed 5 months ago

seonhoon1002 commented 6 months ago

Hello, I have a question regarding the order of the Argoverse validation dataset

When I visualize the Argo dataset processed by your code, the order seems shuffled(other datasets are not shuffled)

Below is a visualization result of Argoverse(of course, I check the shuffle args in dataset loader)

Below is a visualization result of Waymo waymo

In your case, Is Argoverse Dataset shuffled? So it is a natural phenomenon when we get the argoverse dataset by argo2 library?

Abyssaledge commented 6 months ago

I didn't shuffle it on purpose. I believe you can get the ordered results by sorting the resulted pkl file.

seonhoon1002 commented 6 months ago

In Waymo and Nuscense case, their order is correct, yes, so that's why we can use "sweep".

So I have a question about this: does your data processing process involve sorting the order in Waymo or Nuscense case?