🌱 Challenge Chat: How are you guys getting score...

Microsoft Rice Disease Classification Challenge

Helping Egypt

$3 000 USD

Challenge completed ~3 years ago

Skills you will learn

Classification

Computer Vision

916 joined

260 active

Info Data Chat Leaderboard

Start

May 13, 22

Aug 14, 22

Reveal

Aug 14, 22

Koleshjr

Multimedia university of kenya

How are you guys getting scores below 0.10?

Help · 16 Jul 2022, 11:06 · 9

I don't know if i am missing something but i have a good local score of less than 0.10 but a bad lb?

Discussion 9 answers

DoubleAgent

Carnegie mellon university africa

Local cross validation score of less than 0.1?!

16 Jul 2022, 11:13

Upvotes 1

Koleshjr

Multimedia university of kenya

yes whats yours?

replied to DoubleAgent16 Jul 2022, 11:47

Upvotes 0

DoubleAgent

Carnegie mellon university africa

For my current lb score, I have a mean CV score of ~0.094 and yours?!

replied to Koleshjr16 Jul 2022, 15:44

Upvotes 0

DJOE

Ecole Supérieure Privée d'Ingénierie et de Technologies - ESPRIT

Trust your cv dont give attention to the leader board.

16 Jul 2022, 11:46

Upvotes 2

Koleshjr

Multimedia university of kenya

Thank you

replied to DJOE16 Jul 2022, 11:47

Upvotes 1

Koleshjr

Multimedia university of kenya

0.07 but i am not using folds cv, im using fastai stratified splitter that behaves like train_test_split

16 Jul 2022, 16:36

Upvotes 1

DJOE

Ecole Supérieure Privée d'Ingénierie et de Technologies - ESPRIT

I use stratifiedkfold cv with 5 folds. cv 0.10 and lb 0.091.

Try also test time augmentation, it improves results.

replied to Koleshjr17 Jul 2022, 00:34

Upvotes 1

Koleshjr

Multimedia university of kenya

Thank you

replied to DJOE17 Jul 2022, 07:32

Upvotes 0

rockerritesh

It's painfull!!

29 Jul 2022, 18:22

Upvotes 0

Join the largest network for
data scientists and AI builders

About FAQs

Status