search

OpenDriveLab / ST-P3

[ECCV 2022] ST-P3, an end-to-end vision-based autonomous driving framework via spatial-temporal feature learning.
Apache License 2.0
315 stars 38 forks source link

When will you make deep datasets public? #17

Open xiaowhite-22 opened 1 year ago

ilnehc commented 1 year ago

@xiaowhite-22 Thanks for your interest. Due to the huge size (~1T), we did not figure out a great platform to hold the dataset. If you want that, maybe we can find a way to share it with you.

xiaowhite-22 commented 1 year ago

I just want the depth information label for the keyframe, which is 1T?

ilnehc commented 1 year ago

The depth label for supervision is the dense absolute value (float32). Each generated original file is about 49001600/1024/1024=5.5Mb.