This hackathons is only open to students. Double check the event page for more information as this may mean only those from a particular university/country are eligible.
Event Type
online
94
Participants
₹40,000
Prize Pool
8
Est. Projects
Organizers
Alex Johnson
alex@example.org
Jamie Rivera
jamie@example.org
Sam Chen
sam@example.org
Quality Score
Quality Score
72/100
High confidence
Organiser16/20
Event Maturity14/20
Sponsors18/25
Participants12/20
Overview:
Can you tell a chihuahua from a muffin? Put your AI skills to the test in this exciting data-centric AI challenge, conducted in collaboration with 3LC.
Unlike traditional competitions where you tweak models, here your focus is different—you’ll work with a fixed ResNet-18 model and improve performance by making your data smarter, cleaner, and more meaningful.
You’ll begin with a small labeled dataset and a large pool of unlabeled images, and use the powerful 3LC platform to analyze, label, and refine your data step-by-step. Your ultimate goal? Achieve the highest accuracy on a hidden test set through smart data curation and intelligent labeling strategies.
Dataset Details:
Training Set
100 labeled images (50 per class)
3,579 unlabeled images
Validation Set
1,000 labeled images (balanced)
Used for monitoring model performance
Test Set
1,184 images (labels hidden)
Used for leaderboard evaluation
Process & Guidelines:
Participants will follow an iterative data-centric workflow:
TrainTrain the ResNet-18 model using the available labeled dataset
AnalyzeUse the 3LC Dashboard to:
Visualize embeddings
Identify misclassified and uncertain samples
Understand model weaknesses
Improve Data:
Label selected samples from the unlabeled pool
Correct incorrect labels
Use sample weights to include/exclude data
Retrain: Retrain the model using the updated dataset
Submit: Generate predictions on the test set and submit to Kaggle
Iterate: Repeat the process to continuously improve performance and climb the leaderboard
Rules:
Only the ResNet-18 architecture is allowed (no modifications or alternative models)
Only the provided dataset can be used (no external data allowed)
ImageNet pretrained weights are allowed
Use of the 3LC platform is mandatory
All submissions must follow the Kaggle submission format
Participants must adhere to fair competition practices
Any form of data leakage, misuse of test data, or rule violation will result in disqualification
Stay in the loop with all event updates, important announcements, deadlines, and exclusive information. Make sure to join the WhatsApp group so you don’t miss anything essential throughout the event!
Operations12/15
Why this score
Strong organiser track record
Returning event
Well-sponsored
Missing data
Prize details
Code of conduct
Chihuahua vs Muffin: Data-Centric AI Challenge (in collaboration with 3LC) | Hackathon Radar