-
Notifications
You must be signed in to change notification settings - Fork 0
/
build_dataset.py
159 lines (128 loc) · 5.23 KB
/
build_dataset.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
"""Split the MUMU dataset into train/dev/test sets.
The MuMu dataset comes in the following format:
AMAZON_ID.jpg
...
Original images have various sizes at or below (300, 300).
"""
import argparse
import json
import csv
import random
import os
import shutil
import urllib.request as req
import numpy as np
from PIL import Image
from tqdm import tqdm
parser = argparse.ArgumentParser()
parser.add_argument('--data_dir', default='data/MUMU',
help="Directory with MUMU album images")
parser.add_argument('--output_dir', default='data/300x300_MUMU',
help="Where to write the new data")
parser.add_argument('--mumu_metadata', default='data/amazon_metadata_MuMu.json',
help="File with MuMu metadata")
parser.add_argument('--data_labels', default='data/MuMu_dataset_multi-label.csv',
help="File with MuMu album genres")
def create_dir(dir_name):
"""Create directory with check not to overwrite existing directory"""
if not os.path.exists(dir_name):
os.mkdir(dir_name)
else:
print("Warning: dir {} already exists".format(dir_name))
def create_with_overwrite(dir_name):
"""Create new directory or overwrite existing dir if it exists."""
if os.path.exists(dir_name):
shutil.rmtree(dir_name)
os.mkdir(dir_name)
def get_image_urls(mumu_metadata):
"""Retrieve all urls for album cover images in the MuMu dataset"""
with open(mumu_metadata) as f:
data = json.load(f)
img_urls = dict()
for entry in data:
img_url = entry['imUrl']
img_id = entry['amazon_id']
img_urls[img_id] = img_url
return img_urls
def download_images(img_urls, data_dir):
"""Download album cover images from the MuMu dataset to `data_dir`"""
create_dir(data_dir)
for img_id, img_url in tqdm(img_urls.items()):
img_format = img_url[-4:]
img_file = os.path.join(data_dir, img_id + img_format)
req.urlretrieve(img_url, img_file)
def generate_splits(filenames):
"""Generate 80/10/10 train/dev/test splits for data"""
filenames.sort()
random.seed(230)
random.shuffle(filenames)
split_1 = int(0.8 * len(filenames))
split_2 = int(0.9 * len(filenames))
filenames_train = filenames[:split_1]
filenames_dev = filenames[split_1:split_2]
filenames_test = filenames[split_2:]
splits = {'train': filenames_train,
'dev': filenames_dev,
'test': filenames_test}
return splits
def get_album_genres(f_genres):
"""Retrieve genres for each album in the MuMu dataset"""
album_genres = dict()
with open(f_genres) as f:
data = csv.reader(f)
next(data) # skip header
for row in data:
img_id = row[0] # amazon_id
genres = list(row[5].split(','))
album_genres[img_id] = genres
return album_genres
def generate_labels(filenames, album_genres, output_dir, split):
"""Convert genres for each album to binary vector label and save to `output_dir`"""
genre_list = list(set([genre for genres in album_genres.values() for genre in genres]))
genre_list.sort()
labels = []
files = []
for f in filenames:
img_id = f.split('/')[-1][:-4]
genres = album_genres[img_id]
album_label = [1 if g in genres else 0 for g in genre_list]
if 1 not in album_label:
print(album_label)
print(genres)
labels.append(album_label)
files.append(f)
output_file = os.path.join(output_dir, 'y_' + split + '.npy')
np.save(output_file, np.array(labels))
return files
def save(filenames, output_dir):
"""Save the images contained in `filenames` to the `output_dir`"""
for filename in tqdm(filenames):
image = Image.open(filename)
image.save(os.path.join(output_dir, filename.split('/')[-1]))
if __name__ == '__main__':
args = parser.parse_args()
# Scrape album artwork if not already present
if not os.path.isdir(args.data_dir):
print("Dataset at {} not found. Scraping images now.".format(args.data_dir))
img_urls = get_image_urls(args.mumu_metadata)
download_images(img_urls, args.data_dir)
# Get the filenames in the data directory
filenames = [os.path.join(args.data_dir, f) for f in os.listdir(args.data_dir) if f.endswith('.jpg')]
# Split into 80%/10%/10% train/dev/test
splits = generate_splits(filenames)
# Create new output data directory
create_dir(args.output_dir)
# Preprocess train, dev and test images
album_genres = get_album_genres(args.data_labels)
for split, files in splits.items():
dir_split = os.path.join(args.output_dir, split)
create_with_overwrite(dir_split)
output_dir_images = os.path.join(dir_split, 'images')
create_with_overwrite(output_dir_images)
output_dir_genres = os.path.join(dir_split, 'genres')
create_with_overwrite(output_dir_genres)
print("Generating {} labels, saving to {}".format(split, output_dir_genres))
files_present = generate_labels(files, album_genres, output_dir_genres, split)
print("Processing {} data, saving to {}".format(split, output_dir_images))
save(files_present, output_dir_images)
print("Done building dataset")