AgriFieldNet India Challenge
Can you detect crop types in a class-imbalanced satellite image dataset?
$10 000 USD
~1 month to go
78 active · 354 enrolled
Training data!
Data · 16 Sep 2022, 10:24 · 3

Hello, I'm wondering as to why training data is not suited for convolutional neural networks!. Is the challenge only for to non-CNN models?

Discussion 3 answers

Dear Zindi

As per data set of "AgriFieldNet Competition Dataset" the crops given to map are;

1 - Wheat 2 - Mustard 3 - Lentil 4 - No crop/Fallow 5 - Green pea 6 - Sugarcane 8 - Garlic 9 - Maize 13 - Gram 14 - Coriander 15 - Potato 16 - Bersem 36 - Rice.

These crops will have very high spectral,overlap, cannot be mapped with single data data sets as provided, without incorporating phenology of these crops.

We have experimented with large number of crops using sentient-2 as well as PlanetScope temporal data sets.


i guess because of the data size. I don't think its more suitable to train on CNN due to data size. But that's just my opinion!

You have free to make a choice regarding how to tackle the challenge. So, CNN is not excluded if you want to use