Zindi
CompeteJobsLearnChatLeaderboardFor Organisations
PartnersMeet the teamPressCase studiesAI4EAC
Primary competition visual

INEGI UN-GGIM Human Settlement Detection Challenge by ITU

Helping Mexico
1 000 CHF
Completed (~2 years ago)
Classification
213 joined
84 active
InfoDataChatLeaderboard
Starti
02 Sep 24
Closei
28 Sep 24
Reveali
10 Oct 24
User avatar
ML_Wizzard
Nasarawa State University
BASELINE NOTEBOOK
Notebooks Ā· 8 Sep 2024, 09:25 Ā· 10

I've created a baseline for the INEGI-GCIM Human Settlement Detection Challenge by ITU You can find the code at mubrij/INEGI-GCIM-Human-Settlement-Detection-Challenge-by-ITU (github.com). Please note that your results using this code may vary slightly, but my solution achieved an LB score of 0.504...

Discussion 10 answers
User avatar
nymfree
8 Sep 2024, 10:28

Thanks for sharing.

so we are in the same predicament in that we achieve an LB score that is more or less equivalent to random guessing. tried with several models that have good AUC score but LB == random guess. what are we missing?

Upvotes 1
User avatar
ML_Wizzard
11 Sep 2024, 08:39replied to nymfree

@nymfree I haven't looked deeper into the dataset yet because I was too busy, but from what I observed when creating my baseline, it seems that a machine learning model may not accurately predict well in this case. It might be better to try using a deep learning approach to capture the complexity.

Upvotes 0
User avatar
ahmadmwali
10 Sep 2024, 10:00

Thanks for sharing. ML Wizzard and nymfree, the lower LB score is probably a result of overfitting. It will be improved if you included regularization or to make the model simpler. However, since this is an image classification task, I think it's better solved using CNN as it works really well. Thanks.

Upvotes 0
User avatar
nymfree
10 Sep 2024, 13:22replied to ahmadmwali

Thanks for your insight. I will double check the data and my pipeline.

Upvotes 0
User avatar
ML_Wizzard
11 Sep 2024, 08:41replied to ahmadmwali

Sure @ahmadmwali.

Upvotes 0
User avatar
nymfree
10 Sep 2024, 19:51

Figured it out. you have to use the provided id_map file

Upvotes 1
User avatar
ML_Wizzard
11 Sep 2024, 08:40replied to nymfree

@nymfree How did you use the id_map file ??

Upvotes 0
User avatar
nymfree
11 Sep 2024, 08:55replied to ML_Wizzard

Basically after making predictions, assuming all your predictions are in a list, do:

df = pd.read_csv('id_map.csv')
my_dict = {}
for i,row in df.iterrows():
    my_dict[row['id']] = row['ID']
    
sub = pd.read_csv('SampleSubmission.csv')
for i in range(len(sub)):
    sub.at[i,'class'] = test_predictions[my_dict[sub.at[i,'id']]]

Upvotes 3
User avatar
ML_Wizzard
11 Sep 2024, 11:25replied to nymfree

Thank you @nymfree.

Upvotes 0
User avatar
Acalo
12 Sep 2024, 19:15replied to nymfree

thanks mate @nymfree

Upvotes 0

Join the largest network for
data scientists and AI builders

AboutFAQsPrivacy PolicyTerms of UseRules
Status
Ā© Zindi 2026