Primary competition visual

Amini Cocoa Contamination Challenge

Helping Ghana
$7 000 USD
Completed (11 months ago)
Computer Vision
Object Detection
928 joined
255 active
Starti
Feb 14, 25
Closei
May 11, 25
Reveali
May 12, 25
User avatar
nymfree
800 club: what models are you using?
Help · 25 Apr 2025, 14:40 · 15

Cannot break the 0.76 with yolo11 and various pre-/post-processing. what is the secret? Cannot get to reliably train any mmdet model

Discussion 15 answers
User avatar
CodeJoe

I haven't reached the 80 score yet but you can literally reach 78 easily by:

Perhaps try kfolds or reduce the test size if you are using yolo11.

Selecting a different model variant maybe (s, m, or even l)

Tune hyperparameters especially epochs (30-100), batch size.

For post processing the normal way, reduced confidence level (0 or 0.001 - might bring false positives though) , tuned IOU and image size.

For preprocessing, I didn't touch anything😅. I let yolo do that for me.

25 Apr 2025, 14:49
Upvotes 1

Does size L meet the requirements? I'm even doubtful about size M

User avatar
nymfree

Thanks. I will restart some experiments.

Size L is pretty quick to train.

User avatar
CodeJoe

Thats a very good question.

Your solutions for this challenge must be able to function in a resource-limited setting i.e. it should run on a low-resource smartphone. As such, we are imposing the following restrictions on resources: T4 GPU, maximum 9h training, maximum 3h inference Model frameworks must be appropriate for use on edge devices (e.g. ONNX, TensorFlow Lite)

I think yolo11l and m fits in well for this description. It can be used on onnx and isn't that heavy around 50mb. If you break the restrictions that's when your model doesn't fit the description.

ok !! Have you applied class balancing?

User avatar
CodeJoe

That's the next thing I'm going to do. Honestly I thought the classes are well balanced so I didn't do any sort of class balancing. But now I'm suffering to improve the score, so yeah why not 😅

User avatar
CodeJoe

Did it work out for you?

no, I haven’t balanced them yet either.

User avatar
CodeJoe

@nymfree are you still suffering to break the score?

User avatar
nymfree

still struggling. I see that you guys broke through.

User avatar
CodeJoe

Yes we really struggled to break the score. Did you try KFolds? Use a large number of folds and see whether it works.

User avatar
nymfree

tried kfolds and tuning some hyperparameters. can improve local score but not LB. Maybe my splits are not quite right.

User avatar
CodeJoe

What's your CV? Maybe there might be data leakage. Some models perform better than other models though. I know there is some kind of gap when I also post my submissions.

User avatar
nymfree

My splits were not quite right. I was doing something dumb. Re-running my experiments. Thanks @CodeJoe, I was about to give up on this competition.

User avatar
Koleshjr
Multimedia university of kenya

Yolo can get you 0.800

25 Apr 2025, 19:03
Upvotes 1