Individual submission with a non-unique algorithm?

Good afternoon nuScenes team,

Are individuals allowed to make submissions to the challenges with an algorithm based on somebody else’s work?

I’d like to make a submission and I work a full-time 50+ hours/week job, the chance that I can come up with a completely new and unique algorithm myself working ~10 hours/week outside of work in a handful of months is essentially none. However I have some familiarity with the dataset and 3d object detection and agent prediction algorithms, and I could probably make a decent submission based largely on algorithms that already exist in my available time. Is this allowed?

If so, more specifically, I see on the page for the challenge https://www.nuscenes.org/prediction?externalData=all&mapData=all&modalities=Any that the following information is required:

-Team name
-Method name
-Authors
-Affiliations
-Method description (5+ sentences)
-Project URL
-Paper URL
-FPS in Hz (and the hardware used to measure it)

For Team name should I list my name? For Method name, should I list the method on which my submission is based? What if I make substantial modifications/additions? For Authors what if I haven’t authored a paper pertaining to my submission? What if I don’t have any Affiliations? Etc. for the remaining fields.

If you choose to not allow individual submitters that would be a real shame as that would disqualify anyone from participating who is not affiliated with a research organization and who can dedicate themselves to a submission on a full-time or near full-time basis, which is potentially a large number of people.

Whatever your response, thanks for your great work with this dataset, I’m enjoying it very much.

-Chris Dahms

Hi Chris,

no worries, you can definitely submit. We often have teams composed of a single student or other professional. The team name you can make up yourself. Likewise Method name is also up to you, either you can indicate that this is a modification of something existing or come up with a completely new name if you think it’s justified. Author will be just you. Paper is optional.
These days the complexity of leading methods is growing quickly and almost all new methods in the field are based on someone else’s work. So don’t worry about your modifications being too minor to be “unique”. Most breakthrough papers/works start like that.

2 Likes