Ranked 12 in Ens Mauna Kea Challenge Didn't had time to do pseudo labelling, TTA either, so there's at least two ways of improvement, I have others but it's unsure.
Few links that inspired me:
- https://www.dlology.com/blog/multi-class-classification-with-focal-loss-for-imbalanced-datasets/
- https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/77322#latest-460258
- https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/77731#latest-470101
- https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/77320#latest-469487
- https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/78109#458525
- https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/77251#latest-458255
- https://www.kaggle.com/c/histopathologic-cancer-detection/discussion/87400#latest-512392
- https://www.kaggle.com/c/histopathologic-cancer-detection/discussion/87397#latest-515386
- https://www.kaggle.com/c/histopathologic-cancer-detection/discussion/84790#latest-503843
- https://www.kaggle.com/c/histopathologic-cancer-detection/discussion/83760#latest-554506
- https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/77269#latest-456837
- https://machinelearningmastery.com/how-to-create-a-random-split-cross-validation-and-bagging-ensemble-for-deep-learning-in-keras/
- https://machinelearningmastery.com/stacking-ensemble-for-deep-learning-neural-networks/