- Albumentations
- Pytorch Image Models
- Numpy
- Open-CV
- Pytorch
- SciKit-Learn
Regarding the libraries, all rights belong to their respective owners.
Best Private Score achieved:
2x EfficientNet B0: 0.9200 AUC with 512x512 imagesize No Metadata, TTA and Pseudolabelling was used, ~Top 58% Rank
*TPU code has some issues where model trained with TPU, loaded on CPU for inference gave really bad public LB scores but decent CV score.