Wadhwani AI Bollworm Counting Challenge
'%3e%3ccircle r='20' fill='%23008'/%3e%3ccircle r='17.5' fill='%23fff'/%3e%3ccircle r='3.5' fill='%23008'/%3e%3cg id='d'%3e%3cg id='c'%3e%3cg id='b'%3e%3cg id='a' fill='%23008'%3e%3ccircle r='.875' transform='rotate(7.5 -8.75 133.5)'/%3e%3cpath d='M0 17.5.6 7 0 2l-.6 5L0 17.5z'/%3e%3c/g%3e%3cuse xlink:href='%23a' transform='rotate(15)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(30)'/%3e%3c/g%3e%3cuse xlink:href='%23c' transform='rotate(60)'/%3e%3c/g%3e%3cuse xlink:href='%23d' transform='rotate(120)'/%3e%3cuse xlink:href='%23d' transform='rotate(-120)'/%3e%3c/g%3e%3c/svg%3e)
Helping India
€15 000 EUR
Completed (over 3 years ago)
Object Detection
Computer Vision
832 joined
146 active
Start
Sep 23, 22
Close
Dec 04, 22
Reveal
Dec 04, 22
2nd Place Solution
Platform · 5 Dec 2022, 15:13 · 10
Nothing special , just a yolov5l6 trained for 25 epoch with image size 1536. I split folds by mskf
Part 2 i used a classifier which detects Images with no bollworms , helps reduce False positive.
I used Optuna to obtain the best values for Soft NMS(Special Thanks to @ZFTurbo for his ensemble boxes lib) and thresholds for classifier and yolov5l6.
I used A100 and A6000 for compute
Congratulations for the great performance. Would you be opening your code as well? I would love to look at Optuna part.
I will do in a short time
https://github.com/king398/PestDetectFinal my code
Congratulations with your 2nd! May I ask you that your classifier help you boost your CV? And how much? Thanks you
Yes , classifier helped me boost my cv by a small margin
25 epochs, seems that that took a long time to train?
6 minutes with an a100 per epoch
My solution is quite similar to yours, except for optuna part. :)
Congratulations on a great win, it is impressive. Thanks for the recognition of @ZFTurbo.
Looking Forward to your solution