Challenge Chat: I think April 2022 data is...

GeoAI Challenge for Cropland Mapping by ITU

Helping Iran (Islamic Republic of), Sudan
and 1 other country

Iran (Islamic Republic of)
Sudan
Afghanistan
Scroll to see more

$4 000 USD

Challenge completed ~2 years ago

Skills you will learn

Classification

362 joined

74 active

Info Data Chat Leaderboard

Start

Jun 30, 23

Oct 05, 23

Reveal

Oct 05, 23

Rakesh_Jarupula

National Institute of Technology Silchar

I think April 2022 data is not enough for AFG region

Data · 29 Sep 2023, 05:00 · 5

Hello @Lolletti

For the afghanisthan region, training data from April may not be sufficient (atleast for me. Getting ~80% accuracy). This makes sense because in this period most of the crops are harvested, which doesn't help in classification.

And are the labels are credible for the afg region for the train and test period. Because there is huge perfomance gap with respect to other regions.

If the labels were collected during different time periods....then the modelling only for April is not useful idea. Same is true for the test dataset.

I was wondering how the top teams are approaching the AFG region modelling.

Thanks

Discussion 5 answers

enigmatic

I have exactly 80% accuracy for Afg too, difficult to go higher than that

29 Sep 2023, 06:11

Upvotes 0

Lolletti

Thanks for your query. I confirm that data were collected in April 2022 and are based on visual interpretation of VHR images. Participants are allowed to increase the test dataset with open source data. Considering the challenge represented by the limited time window, 80% of accuracy isn't that bad in my personal opinion.

29 Sep 2023, 07:50

Upvotes 1