below you can find my code solution:https://github.com/TAUIL-Abd-Elilah/Competitions/blob/main/Zindi/CGIAR%20Eyes%20on%20the%20Ground%20Challenge/CGIAR-Eyes-on-... See more

Anyone who used Fastai, and is In the Top 20?? Would you be so kind to share the solution ??

I absolutely do not trust that the top 1 team can achieve both public and private score that are gap to the rest. Clearly they used data leakage in some way. Be... See more

As I see, the final leaderboard has been revealed, but I see that there is a 1st team that has a very different score, so I'm sure they used type of damage? Tha... See more

Training and inference source code: https://github.com/quanvuhust/cgiar-eyes-on-the-ground-challengeSPLIT DATAI split train data to 5 fold using group kfold. Gr... See more

what are the tricks used by top teams to get such great score .I have used lots of different methods but only reached 8.91 /9.99 (public/private) without any le... See more

Thanks to the organizers of the competition and to all the participants who shared their experiences in the discussions. I've decided to write a brief report on... See more

Hi. For my last 5 submissions, my leaderboard score has been giving the exact same value despite the fact that I made them using different architectures. I know... See more

I realized I am required to use root mean squared error for my loss fuction, but RMSE does not exist in tf.keras.losses. I have trained a model using MSE, is th... See more

@zindi i was unable to either team up or send a team up request. This thing is becoming rampant nowadays, is there anything wrong or what ? I am trying to addre... See more

We can label samples in test set to zeros if type of damage is non DR. And using them for training. Is this allowed?

T

I think earlier training time limit was 9 hours with p100 gpus, and now changed to 8 hours. Is it updated or i just misread previously. As google colabs has t4... See more

What's the best way to incorporate growth stage? I've tried postprocessing with growth stage + predictions, this only gave modest improvements however. Has anyo... See more

I get this error while trying to download the data

Can we use the 'season' column for prediction? According to the official description of the application scenario, it seems that the model will be applied to dif... See more

Let's share our CV vs LB scores. I'll start:Model: Resnet34Data augmentation: Horizontal rotation, BrightnessCV: 8.79LB: 16.09Anyone seeing a similar gap betwee... See more

What are the names/types of the crops available in the dataset? Are they all of the same type?

'We therefore invite you to improve on the existing solutions, and improve the accuracy at which your model predicts drought damage across multiple seasons.' Ha... See more

The description of the target value extent says it is in this range of 0-100 with increments of 10%. ## Ancillary data For each growth stage the damage types an... See more

Well am downloading data from Colab but its taking long to download the zip files though i have the test and train folders but they are not downloadable until... See more

Considering damage type is not used in train and test data.Sorry for the naive question: Are you solving this as a regression problem or a classifcation problem... See more

...

@Zindi it's clear now that we're not allowed to use the "type of damage" column. But how we could know if we can trust the LB to monitor our progress and see if... See more

Hello, I'm trying to get the images in the submissionfile.csv so I can make a submission. Using the ID column in the submission file, I tried looking for it's e... See more