Kaggle Competition: Human Protein Atlas Classification

This is a very interesting and challenging problem due to following reasons:

1. Extremely imbalanced class sizes
2. Multi-class classification (28 classes)
3. Use of external data
4. Volume of data (external data itself in 190 GB of 2048*2048 images)

The code and solution here:

• Github > AksChougule > kaggle-competitions > Human-Protein-Atlas