Notebooks and Code for the 2018 Data Science Bowl

In the 2018 Data Science Bowl, a Kaggle competition, the competitors were provided with microscopic images of nuclei and were tasked to identify them. This repository contains my notebooks and code written for this competition.

Data Source

Data are provided by the competition's host and can be downloaded here. To run the notebooks, unzip the training data at data/train/ and test data at data/test.

Performance (mean IoU, Kaggle LB score)

U-Net Baseline - 0.310
U-Net Benchmark - 0.355