starter-challenge-data-2020

Info

This starter contains:

• helper functions to convert to RLE (Run-length encoding) and back
• a scoring function for Dice Score
• a starter project using Keras to generate predictions

Requirements

• Python 3.5+
• numpy==1.16.1
• opencv-python==3.4.5.20
• pandas==0.23.4

Helpers

make_rle_submission.py

Usage:

python make_rle_submission.py --mask_folder ../data_challenge/output --output_path submission.csv --sample_csv_path sample_submission.csv

Note that the mask_folder must only contain .png files with binary masks (values in [0, 1] or [0, 255])

score_submission.py

Usage:

python score_submission.py --submission_csv_path submission.csv --ground_truth_csv_path ./ground_truth.csv

Util functions

RLE decode

from utils improt rle_encode, rle_to_string

rle_runs = rle_encode(mask)
rle_string = rle_to_string(rle_runs)

RLE encode

from utils improt rle_encode, rle_to_string

size = (720, 1280)
mask = rle_decode(rle_string, (720, 1280))

Dice score

from utils import dice_score

score = dice_score(grount_truth, prediction)

Starter project

Heavily inspired from the Kaggle Carvana's third place solution, it is a simple Keras sample allowing competitors to get started. Get started with benchmark.ipynb.