A solution for https://www.kaggle.com/c/carvana-image-masking-challenge
- Strided and transposed convolutions
- Added dilated convs
Notebook might have a lot of images so it takes a while to load.
I was changing the logging during training, so there's some gaps in the data
Accuracy
Loss
