-
Notifications
You must be signed in to change notification settings - Fork 42
/
post_processing.py
169 lines (128 loc) · 5.39 KB
/
post_processing.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
import argparse
import tqdm
import scipy.misc as misc
import pandas as pd
from collections import defaultdict
def process_duplicate_ids(df_submission, df_duplicate):
duplicate_ids_dict = defaultdict(list)
for _, row in df_duplicate.iterrows():
ids = row['SubId'].split(' ')
if len(ids) == 2:
duplicate_ids_dict[ids[0]].append(ids[1])
duplicate_ids_dict[ids[1]].append(ids[0])
else:
duplicate_ids_dict[ids[0]].extend([ids[1], ids[2]])
duplicate_ids_dict[ids[1]].extend([ids[0], ids[2]])
duplicate_ids_dict[ids[2]].extend([ids[0], ids[1]])
records = []
for image_filenae, row in tqdm.tqdm(df_submission.iterrows(), total=len(df_submission)):
ids = row['Id'].split(' ')
if ids[0] not in duplicate_ids_dict:
records.append((image_filenae, row['Id']))
continue
duplicate_ids = duplicate_ids_dict[ids[0]]
new_ids = ids[:1] + duplicate_ids
ids = new_ids + [v for v in ids if v not in new_ids]
ids = ids[:5]
assert len(ids) == len(set(ids))
records.append((image_filenae, ' '.join(ids)))
return pd.DataFrame.from_records(records, columns=['Image', 'Id'], index='Image')
def get_image_size(filepath):
return misc.imread(filepath).shape[:2]
def get_image_sizes(filenames, images_dir):
ret = set()
for filename in filenames:
filepath = os.path.join(images_dir, filename)
ret.add(get_image_size(filepath))
return ret
def build_size_dict(df_group, df_train):
leak_id_set = set()
ret_dict = {}
for _, row in df_group.iterrows():
id_list = row['SubId'].split(' ')
size_dict = {}
for id_str in id_list:
filenames = df_train[df_train['Id'] == id_str].index
size_dict[id_str] = get_image_sizes(filenames, os.path.join('data', 'train'))
leak_id_set.add(id_str)
ret_dict[tuple(list(sorted(id_list)))] = size_dict
return ret_dict, leak_id_set
def update_using_image_size_leak(df_submission, df_duplicate, df_train):
size_dict, leak_id_set = build_size_dict(df_duplicate, df_train)
records = []
for image_filename, row in tqdm.tqdm(df_submission.iterrows(), total=len(df_submission)):
ids = row['Id'].split(' ')
if ids[0] not in leak_id_set:
records.append((image_filename, row['Id']))
continue
leak_dict = None
for key, value in size_dict.items():
if ids[0] in key:
leak_dict = value
break
cur_size = get_image_size(os.path.join('data', 'test', image_filename))
target_id = None
for key, value in leak_dict.items():
if cur_size in value:
if target_id is not None:
target_id = None
break
target_id = key
if target_id is None:
records.append((image_filename, row['Id']))
continue
new_ids = [target_id] + [v for v in ids if v != target_id]
assert len(new_ids) == 5
records.append((image_filename, ' '.join(new_ids)))
return pd.DataFrame.from_records(records, columns=['Image', 'Id'], index='Image')
def update_using_data_leak(df_submission, df_leak):
changed = []
records = []
for image_filename, row in tqdm.tqdm(df_submission.iterrows(), total=len(df_submission)):
ids = row['Id'].split(' ')
if image_filename not in df_leak.index:
records.append((image_filename, row['Id']))
continue
leak_id = df_leak.loc[image_filename]['Id']
top1_id = ids[0]
if leak_id == top1_id:
records.append((image_filename, row['Id']))
continue
ids = [Id for Id in ids if Id != leak_id]
ids = [leak_id] + ids
changed.append((image_filename, row['Id'], leak_id))
assert len(set(ids)) == len(ids)
records.append((image_filename, ' '.join(ids)))
return pd.DataFrame.from_records(records, columns=['Image', 'Id'], index='Image')
def parse_args():
description = 'post processing'
parser = argparse.ArgumentParser(description=description)
parser.add_argument('--input_path', dest='input_path',
help='input file path',
default=None, type=str)
parser.add_argument('--output_path', dest='output_path',
help='result file path',
default=None, type=str)
return parser.parse_args()
def main():
import warnings
warnings.filterwarnings("ignore")
print('post processing')
args = parse_args()
print('from', args.input_path)
print('to', args.output_path)
df_submission = pd.read_csv(args.input_path, index_col='Image')
df_leak = pd.read_csv('data/leaks.csv', index_col='Image')
df_train = pd.read_csv('data/train.csv', index_col='Image')
df_duplicate = pd.read_csv('data/duplicate_ids.csv')
df_submission = process_duplicate_ids(df_submission, df_duplicate)
df_submission = update_using_image_size_leak(df_submission, df_duplicate, df_train)
df_submission = update_using_data_leak(df_submission, df_leak)
df_submission.to_csv(args.output_path)
print('success!')
if __name__ == '__main__':
main()