Solution based on U-Net for the Kaggle Carvana Image Masking Challenge
This solution was based on Heng CherKeng's code for PyTorch. I kindly thank him for sharing his work. 128x128, 256x256, 512x512 and 1024x1024 U-nets are implemented. Public LB scores for each U-net are:
U-net | LB score |
---|---|
128x128 | 0.990 |
256x256 | 0.992 |
512x512 | 0.995 |
1024x1024 | 0.996 |
test_submit_multithreaded.py
to try it.Place 'train', 'train_masks' and 'test' data folders in the 'input' folder.
Convert training masks to .png format. You can do this with:
mogrify -format png *.gif
in the 'train_masks' data folder.
Run python train.py
to train the model.
Run python test_submit.py
to make predictions on test data and generate submission.