NASA Harvest Field Boundary Detection Challenge
Can you detect field boundaries for Rwandan smallholder farmers in a satellite image dataset?
$5 000 USD
Ended 22 days ago
183 active · 724 enrolled
Earth Observation
when you go to view old submissions ...
Data · 27 Feb 2023, 09:42 · 6

congrats to the winners, but I'm curious if anyone has the same :D

Discussion 6 answers

Wow @Riziko - you actually won by a wide margin, nobody knows it yet.

Just refresh that page, I know on Zindi you have to refresh often. If it stays like that ... well ... then you have the biggest GPU of 'em all ...

27 Feb 2023, 10:45
Upvotes 0

however I don't know how to reproduce the old code, so it doesn't matter. it's just funny, maybe someone did even better but they don't know it :D

27 Feb 2023, 12:08
Upvotes 0

Hello @Riziko,

hrere is what i suggest :

download the 0.44 score submission file and resubmit it .

  • if it gives you the same score,Hard luck and Great Job
  • if it didn't , here is a possible explanation :

  1. Keep in mind that ZINDI shows always your best submission score on public LB .
  2. there was a metric problem in this competition (before 3 months).and as i remember that there was 1 , 0.999 etc ... as public scores . So your 0.44 score must be with the old metric.
  3. My guess is that ZINDI public / private scores are static , i mean that if you have a submition with a wrong metric_1 your scores will be saved in the database , and they will not be updated when zindi change the metric .
  4. so the 0.44 scores are with the old metric which is recall scores.

27 Feb 2023, 12:47
Upvotes 1

Yes, very possible that this might be an old score while using an old metric. I could see also that both csv files were uploaded 3 months ago.

Now that you mention it, it's very likely. I was deceived that the platform gave a rating with the old metric :)