def check(data_dir):
imgs = get_files(data_dir, 'jpg')
xmls = get_files(data_dir, 'xml')
xmls_set = set([xml.replace(".xml", "") for xml in xmls])
imgs_set = set([img.replace(".jpg", "") for img in imgs])
diff = imgs_set.difference(xmls_set)
print(len(diff))
def find_invalid_jpgs(data_path):
print("start to check!")
imgs = get_files(data_path, 'jpg')
print("there are %d files !" % len(imgs))
for img in imgs:
if not is_valid_jpg(img):
print(img)
def rename_file(data_dir):
xmls = get_files(data_dir, 'xml')
print("there are %d xmls" % len(xmls))
for xml in xmls:
tree = ET.ElementTree(file=xml)
name = os.path.split(xml)[-1].replace(".xml", '.jpg')
print(name)
tree.find('./filename').text = name
tree.write(xml)
def get_shape(img_path):
imgs = get_files(img_path, 'jpg')
print("there are %d imgages in total !"%len(imgs))
#img_shape = {}
for img in imgs:
height, width, _ = cv2.imread(img).shape
print(img)
xml_path = img.replace('jpg','xml')
tree = ET.ElementTree(file=xml_path)
tree.find('./size/width').text = str(int(width))
tree.find('./size/height').text = str(int(height))
tree.write(xml_path)
def find_invalid_xmls(data_dir):
xmls = get_files(data_dir,suffix='xml')
print("there are %d files need to check"%len(xmls))
#valid_xmls = []
for xml in xmls:
invalid_bboxes,info = check_bboxes(xml)
if len(invalid_bboxes) > 0:
print(xml)
print(info)
#print(len(invalid_bboxes))
for bndbox in invalid_bboxes:
print(format_bbox(bndbox))
def get_stats(path):
xmls = get_files(path)
labels = {}
for xml in xmls:
tree = ET.ElementTree(file=xml)
names = [k.text for k in tree.findall("/object/name")]
for name in names:
if name not in labels:
labels[name] = 1
else:
labels[name] += 1
for key, value in labels.items():
print(key, "\t", value)
def get_stats_of_bboxes(xml_dir, save_dir):
xmls = get_files(xml_dir, ".xml")
statistics = defaultdict(list)
for xml in xmls:
tree = ET.ElementTree(file=xml)
objs = tree.findall("./object")
for obj in objs:
name = obj.find("./name")
xmin = obj.find('./bndbox/xmin').text
xmax = obj.find("./bndbox/xmax").text
ymin = obj.find("./bndbox/ymin").text
ymax = obj.find('./bndbox/ymax').text
statistics[name].append([int(k) for k in [xmin, xmax, ymin, ymax]])
with open(os.path.join(save_dir, 'statistics.pickle'), 'rb') as f:
pickle.dump(statistics, f)
def get_mean_of_rgb(path, njob=1):
imgs = get_files(path, 'jpg')
nums_img = len(imgs)
block = nums_img // njob
pool = Pool(processes=njob)
res = []
for k in njob:
if k != njob - 1:
#imgs_tmp = nums_img[k*block:(k+1)*block]
res.append(
pool.apply_async(_sum_rgb,
(imgs[k * block:(k + 1) * block], )))
else:
res.append(pool.apply_async(_sum_rgb, (imgs[k * block:], )))
if len(res) != njob:
raise ValueError("some error!")
r = 0
g = 0
b = 0
for k in res:
r += k[0]
g += k[1]
b += k[2]
return r / nums_img, g / nums_img, b / nums_img
def find_invalid_jpgs(data_path):
imgs = get_files(data_path, 'jpg')
for img in imgs:
if not is_valid_jpg(img):
print(img)