sample_submission.py

"""
This file generates the sample_submission.csv baseline. 
Ensure that the 'public_data' is in the same directory 
as this file. 
"""


import pandas as pd
import json 
import numpy as np 

import os



"""
Load the required metadata 
"""
annotation_names = json.load(open(os.path.join('public_data', 'annotations.json')))



"""
Calculate training training class distribution
"""

prior_probs = np.zeros(len(annotation_names))

for ii in xrange(1, 11): 
    df = pd.read_csv(os.path.join('public_data', 'train', str(ii).zfill(5), 'targets.csv'))
    
    non_nans = df[df.isnull().any(axis=1) == False]
    prior_probs += np.asarray(non_nans.mean(axis=0)[annotation_names].tolist())
        
prior_probs /= prior_probs.sum()
prior_probs = prior_probs.tolist()



"""
Generate submission file
"""

se_cols = ['start', 'end']

with open(os.path.join('public_data', 'sample_submission.csv'), 'w') as fil: 
    for te_ind_str in sorted(os.listdir(os.path.join('public_data', 'test'))):
        te_ind = int(te_ind_str)

        meta = json.load(open(os.path.join('public_data', 'test', te_ind_str, 'meta.json')))

        starts = range(meta['end'])
        ends = range(1, meta['end'] + 1)

        for start, end in zip(starts, ends):
            row = [te_ind, start, end] + prior_probs

            fil.write(','.join(map(str, row)))
            fil.write('\n')