Primary competition visual

ICLR Workshop Challenge #1: CGIAR Computer Vision for Crop Disease

Helping East Africa
$5 000 USD
Completed (almost 6 years ago)
Classification
Computer Vision
1050 joined
338 active
Starti
Jan 29, 20
Closei
Mar 28, 20
Reveali
Mar 29, 20
User avatar
ZINDI
Computer vision challenge
Platform · 20 Feb 2020, 07:59 · 4

Dear Zindians,

The ICLR challenges are computer vision challenges, therefore your models should not use any of the metadata provided.

All the best,

The Zindi team

Discussion 4 answers
User avatar
alka

hello, can you please clarify what you mean by metadata ? does it have to do with things like original image size, ... ?

23 Feb 2020, 10:54
Upvotes 0

@alka IMHO image size is also a metadata so it should not be used. It is basically a leak that can boost your score on the Leaderboard but brings no value outside of this dataset. I assume @zindi, CGIAR and ICLR would prefer something that generalizes well and therefore the use of image size should NOT be allowed here.

PS: zindi@zindi.africa is pinged now

User avatar
ZINDI

Image size, aspect ratio, pixel count, etc are not allowed. CGIAR would like a solution that is generalizable.

Can we use apect ratio during resizing? eg turning a 800x400 image into 128x64 then padding to 128x128?