Adaptation for learning of Ensemble Neural Network of the Kaggle competition winner code and solution in a Jupyter Notebook.
Has been added:
-
Weight & Biases to plot the model metrics.
-
Result comparison between Ensemble Neural Networks using the metadata of the images and a Convolutional Neural Network.
The model was trained only on 2 Kfolds, 15 epochs in the preprocessed data of 256x256 size without a resize for the data augmentation to fit the Kaggle limits, modify below that consideration.
The ensemble model result has a score in the kaggle competition of:
- Private: 0.9208 Public : 0.9241
The CNN model result has a score in the kaggle competition of:
- Private: 0.9078 Public : 0.9158
W&B public plots: URL
- Accuracy
- Ensemble Network loss
- CNN loss
- Epochs
- Folds
- This work was inspired by 1st place Melanoma winners public repository :
https://github.com/haqishen/SIIM-ISIC-Melanoma-Classification-1st-Place-Solution