File size: 1,903 Bytes
ebd4e51
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import os
import base64
import numpy as np
import csv
import sys
import zlib
import time
import mmap
import argparse

parser = argparse.ArgumentParser()

# output_dir
parser.add_argument('--downloaded_feats', default='data/bu_data', help='downloaded feature directory')
parser.add_argument('--output_dir', default='data/cocobu', help='output feature files')

args = parser.parse_args()

csv.field_size_limit(sys.maxsize)


FIELDNAMES = ['image_id', 'image_w','image_h','num_boxes', 'boxes', 'features']
infiles = ['trainval/karpathy_test_resnet101_faster_rcnn_genome.tsv',
          'trainval/karpathy_val_resnet101_faster_rcnn_genome.tsv',\
          'trainval/karpathy_train_resnet101_faster_rcnn_genome.tsv.0', \
           'trainval/karpathy_train_resnet101_faster_rcnn_genome.tsv.1']

os.makedirs(args.output_dir+'_att')
os.makedirs(args.output_dir+'_fc')
os.makedirs(args.output_dir+'_box')

for infile in infiles:
    print('Reading ' + infile)
    with open(os.path.join(args.downloaded_feats, infile), "r") as tsv_in_file:
        reader = csv.DictReader(tsv_in_file, delimiter='\t', fieldnames = FIELDNAMES)
        for item in reader:
            item['image_id'] = int(item['image_id'])
            item['num_boxes'] = int(item['num_boxes'])
            for field in ['boxes', 'features']:
                item[field] = np.frombuffer(base64.decodestring(item[field].encode('ascii')), 
                        dtype=np.float32).reshape((item['num_boxes'],-1))
            np.savez_compressed(os.path.join(args.output_dir+'_att', str(item['image_id'])), feat=item['features'])
            np.save(os.path.join(args.output_dir+'_fc', str(item['image_id'])), item['features'].mean(0))
            np.save(os.path.join(args.output_dir+'_box', str(item['image_id'])), item['boxes'])