def create_jpg_folder(root_path_list):
for path, name in image_path_generator(root_path_list):
jpg_folder = os.path.join(path, "jpg")
if not os.path.exists(jpg_folder):
os.mkdir(jpg_folder)
img_ori_path = os.path.join(path, name)
img_target_path = os.path.join(jpg_folder, name)
shutil.move(img_ori_path, img_target_path)
def image_matching_inspect(root_path_list):
image_without_xml_list = []
for path, name in image_path_generator(root_path_list):
image_path = os.path.join(path, name)
xml_path = os.path.join(path, "xml")
xml_name = name.replace('.jpg', '.xml').replace('.JPG', '.xml')
xml_path = os.path.join(xml_path, xml_name)
if not os.path.exists(xml_path):
image_without_xml_list.append(image_path)
return image_without_xml_list
def image_rename(root_path_list):
length = get_image_number(root_path_list)
rename_file_number = 0
for path, name in tqdm(image_path_generator(root_path_list), total=length):
if '.JPG' in name or '.JPEG' in name or '.jpeg' in name:
rename_file_number += 1
file_path = os.path.join(path, name)
file_path_rename = file_path.replace(".JPG", ".jpg").replace('.JPEG', '.jpg').replace('.jpeg', '.jpg')
os.rename(file_path, file_path_rename)
print("There are {}/{} images have be renamed".format(rename_file_number, length))
def postfix_checkup(root_path_list):
postfix_invalid_list = []
length = get_image_number(root_path_list)
for path, name in tqdm(image_path_generator(root_path_list), total=length):
if not name.split('.')[-1].lower() in ['jpg', 'jpeg']:
sub_root_path = path[:path.find(path.split("\\")[-1].split('/')[-1]) - 1]
name = name.split('.')[0]
postfix_invalid_list.append(f'{sub_root_path},{name}')
with open("../result/postfix_invalid_images.txt", 'w', encoding='utf-8') as f:
f.write('\n'.join(postfix_invalid_list))
print("There are {}/{} images have invalid postfix".format(len(postfix_invalid_list), length))
def dup_hash(root_path_list):
length = get_image_number(root_path_list)
hash_dict = {}
print("正在计算图片哈希值……")
for path, filename in tqdm(image_path_generator(root_path_list),
total=length):
img_path = os.path.join(path, filename)
img_hash = image_md5(img_path)
hash_dict.setdefault(img_hash, [])
hash_dict[img_hash].append((path, filename))
with open("../../result/new_dup_hash.json", "w") as f:
json.dump(hash_dict, f, indent=4)
print("总共{}张图片，去重后图片数量为{}".format(length, len(hash_dict)))
def rm_orientation(root_path_list):
length = get_image_number(root_path_list)
for path, name in tqdm(image_path_generator(root_path_list), total=length):
image_path = os.path.join(path, name)
img = Image.open(image_path)
if "exif" in img.info:
exif_dict = piexif.load(img.info["exif"])
if piexif.ImageIFD.Orientation in exif_dict['0th']:
print(image_path)
orientation = exif_dict["0th"].pop(piexif.ImageIFD.Orientation)
exif_dict['Exif'][41729] = b'1'
exif_bytes = piexif.dump(exif_dict)
img.save(image_path, exif=exif_bytes)
def size_checkup_fix(root_path_list, annotations_folder='Annotations', images_folder='JPEGImages'):
length = get_image_number(root_path_list)
wrong_annotations = []
for path, name in tqdm(image_path_generator(root_path_list), total=length):
image_path = os.path.join(path, name)
xml_path = path.replace(images_folder, annotations_folder)
xml_name = name.replace('.jpg', '.xml').replace('.JPG', '.xml')
xml_path = os.path.join(xml_path, xml_name)
with open(xml_path, encoding='utf-8') as f:
tree = ET.parse(f)
tree_root = tree.getroot()
size_element = tree_root.find('size')
xml_width = size_element.find('width').text
xml_height = size_element.find('height').text
with Image.open(image_path) as im:
img_width, img_height = im.size
if not (int(xml_width) == img_width and int(xml_height) == img_height):
size_element.find('width').text = str(img_width)
size_element.find('height').text = str(img_height)
tree.write(xml_path)
with open(xml_path, encoding='utf-8') as f:
tree = ET.parse(f)
tree_root = tree.getroot()
for obj in tree_root.iter('object'):
bndbox_element = obj.find("bndbox")
xmin = int(bndbox_element.find('xmin').text)
xmax = int(bndbox_element.find('xmax').text)
ymin = int(bndbox_element.find('ymin').text)
ymax = int(bndbox_element.find('ymax').text)
if xmin < 0 or xmax > int(xml_width) or ymin < 0 or ymax > int(xml_height):
sub_root_path = path[:path.find(path.split("\\")[-1].split('/')[-1]) - 1]
name = name.split('.')[0]
wrong_annotations.append(f'{sub_root_path},{name}')
break
# 输出统计信息
print(f"{len(wrong_annotations)}/{length} files have wrong annotation")
# 写入文件
with open("../result/select_by_classes.txt", 'w', encoding='utf-8') as f:
f.write('\n'.join(wrong_annotations))
def valid_checkup(root_path_list):
broken_list = []
length = get_image_number(root_path_list)
for path, name in tqdm(image_path_generator(root_path_list), total=length):
image_path = os.path.join(path, name)
try:
with open(image_path, 'rb') as f:
value_buf = f.read()
img_np = np.frombuffer(value_buf, np.uint8)
img = cv.imdecode(img_np, IMREAD_COLOR)
cv.cvtColor(img, cv.COLOR_BGR2RGB, img)
# if img.shape is None:
# sub_root_path = path[:path.find(path.split("\\")[-1].split('/')[-1]) - 1]
# name = name.split('.')[0]
# broken_list.append(f'{sub_root_path},{name}')
except Exception:
sub_root_path = path[:path.find(path.split("\\")[-1].split('/')[-1])-1]
name = name.split('.')[0]
broken_list.append(f'{sub_root_path},{name}')
with open("../result/checkup_broken_images.txt", 'w', encoding='utf-8') as f:
f.write('\n'.join(broken_list))
print("There are {}/{} images broken".format(len(broken_list), length))
def get_image_number(root_path_list):
root_path_list = root_path_list.copy()
return sum(1 for _, _ in image_path_generator(root_path_list))