Created
July 21, 2017 22:15
-
-
Save zhreshold/f33f79342c312bed2fb088a445af0c87 to your computer and use it in GitHub Desktop.
Create toy pascal dataset
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import numpy as np | |
OLD_CLASSES = ['aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', | |
'cat', 'chair', 'cow', 'diningtable', 'dog', 'horse', 'motorbike', | |
'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor'] | |
NEW_CLASSES = ['bicycle', 'bus', 'car', 'cat', 'cow', 'dog', 'horse', 'motorbike', | |
'person', 'train'] | |
OLD_INDICES = [OLD_CLASSES.index(x) for x in NEW_CLASSES] | |
AREA_LIMIT = 0.15 | |
def filter_images(lst_file, new_file): | |
with open(lst_file, 'r') as f: | |
out_list = [] | |
count = 0 | |
for line in f.readlines(): | |
l = line.split() | |
label = l[1:-1] | |
path = l[-1] | |
head_width = int(label[0]) | |
obj_width = int(label[1]) | |
objects = np.reshape(np.array(label[head_width:]).astype(float), (-1, obj_width)) | |
new_label = [] | |
keep = True | |
for k, obj in enumerate(objects): | |
if int(obj[0]) in OLD_INDICES: | |
width = obj[3] - obj[1] | |
height = obj[4] - obj[2] | |
if width * height < AREA_LIMIT: | |
new_label = [] | |
keep = False | |
break | |
else: | |
obj[0] = OLD_INDICES.index(int(obj[0])) | |
new_label.append(obj) | |
if keep and new_label: | |
new_label = np.array(new_label) | |
new_line = [count] + label[:head_width] + objects.ravel().tolist() + [path] | |
new_line = '\t'.join([str(x) for x in new_line]) | |
out_list.append(new_line) | |
count += 1 | |
print(new_file, 'count: ', count) | |
if out_list: | |
with open(new_file, 'w') as f: | |
for line in out_list: | |
f.write(line + '\n') | |
def save_syntxt(fname): | |
with open(fname, 'w') as f: | |
for c in NEW_CLASSES: | |
f.write(c + '\n') | |
if __name__ == '__main__': | |
filter_images('val.lst', 'val2.lst') | |
filter_images('train.lst', 'train2.lst') | |
save_syntxt('names.txt') |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment