This repository has been archived by the owner on May 21, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocess.py
51 lines (41 loc) · 1.78 KB
/
preprocess.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
import codecs
import json
import os
from typing import Dict
from PIL import Image
from annotation_parser import parse_xml
def crop_objects(image_path: str, anno_path: str, output_dir: str) -> None:
image = Image.open(image_path)
annotation = parse_xml(anno_path)
for i, obj in enumerate(annotation['objects']):
obj_name = obj['name']
bndbox = obj['bndbox']
obj_image = image.crop(
(bndbox['xmin'], bndbox['ymin'], bndbox['xmax'], bndbox['ymax']))
category_dir = os.path.join(output_dir, obj_name)
if not os.path.exists(category_dir):
os.makedirs(category_dir, exist_ok=True)
obj_image.save(os.path.join(
category_dir, os.path.split(image_path)[1] + f'_{i}.png'), 'PNG')
def get_categories(anno_path: str) -> Dict[int, str]:
categories: dict[int, str] = {}
index = 0
for file in os.listdir(anno_path):
if file.endswith('.xml'):
annotation = parse_xml(os.path.join('datasets/Annotations', file))
for obj in annotation['objects']:
if obj['name'] not in categories.values():
categories.update({index: obj['name']})
index += 1
return categories
if __name__ == '__main__':
# Crop objects from images
for image_name in os.listdir('datasets/JPEGImages'):
image_path = os.path.join('datasets/JPEGImages', image_name)
anno_path = os.path.join('datasets/Annotations', image_name.replace(
'.jpg', '.xml'))
crop_objects(image_path, anno_path, 'datasets/Cropped')
# Dump categories to json file
categories = get_categories('datasets/Annotations')
with codecs.open('categories.json', 'w', encoding='utf-8') as f:
json.dump(categories, f, indent=4, ensure_ascii=False)