MultiOutput UNet

Overview

Second solution uses U-Net as a base model for customizations. We have added auxiliary outputs, that is nuclei centers and contours, so multi-output U-Net learns three outputs. This solution is based on the previous one and defined in the pipelines.py:L42 (both training and inference).

Parameters that are set will give you approximately 0.371 on the leaderboard (top 8% as of Feb 11th).

Run default experiment

Run command:

$ neptune login
$ neptune send main.py --worker gcp-gpu-large --environment pytorch-0.2.0-gpu-py3 -- train_evaluate_predict_pipeline --pipeline_name unet_multitask

When training is completed, collect Kaggle submit from: /output/dsb/experiments/submission.csv.