一般情況下,自制的voc格式數據集隨着數據量的增加,產生壞數據的概率就會提升,模型往往不可讀這些壞的樣本
人工的檢測又會大大的提升工作量,而且很多壞樣本可視化檢測條件較差,因此需要設計腳本來自動檢測錯誤樣本
以下腳本是我在網上找的,設計非常清晰,可以準確定位出惡劣標註樣本的文件夾/文件名稱
import os
import xml.etree.ElementTree as ET
import numpy as np
# np.set_printoptions(suppress=True, threshold=np.nan)
import matplotlib
from PIL import Image
def parse_obj(xml_path, filename):
tree = ET.parse(xml_path + filename)
objects = []
for obj in tree.findall('object'):
obj_struct = {}
obj_struct['name'] = obj.find('name').text
bbox = obj.find('bndbox')
obj_struct['bbox'] = [int(bbox.find('xmin').text),
int(bbox.find('ymin').text),
int(bbox.find('xmax').text),
int(bbox.find('ymax').text)]
objects.append(obj_struct)
return objects
def read_image(image_path, filename):
im = Image.open(image_path + filename)
W = im.size[0]
H = im.size[1]
area = W * H
im_info = [W, H, area]
return im_info
if __name__ == '__main__':
image_path = '/data/dataset/xxx/VOC2012/JPEGImages/'
xml_path = '/data/dataset/xxx/VOC2012/Annotations/'
filenamess = os.listdir(xml_path)
filenames = []
for name in filenamess:
name = name.replace('.xml', '')
filenames.append(name)
recs = {}
ims_info = {}
obs_shape = {}
classnames = []
num_objs = {}
obj_avg = {}
for i, name in enumerate(filenames):
recs[name] = parse_obj(xml_path, name + '.xml')
ims_info[name] = read_image(image_path, name + '.jpg')
for name in filenames:
im_w = ims_info[name][0]
im_h = ims_info[name][1]
im_area = ims_info[name][2]
for object in recs[name]:
if object['name'] not in num_objs.keys():
num_objs[object['name']] = 1
else:
num_objs[object['name']] += 1
ob_w = object['bbox'][2] - object['bbox'][0]
ob_h = object['bbox'][3] - object['bbox'][1]
ob_area = ob_w * ob_h
if object['bbox'][3] > im_h:
print("這個標籤有問題:{}".format(name))
print("這個目標有問題:{}".format(object['name']))
if object['bbox'][2] > im_w:
print("這個標籤有問題:{}".format(name))
print("這個目標有問題:{}".format(object['name']))
if object['bbox'][1] > im_h:
print("這個標籤有問題:{}".format(name))
print("這個目標有問題:{}".format(object['name']))
if object['bbox'][0] > im_w:
print("這個標籤有問題:{}".format(name))
print("這個目標有問題:{}".format(object['name']))
ob_h = object['bbox'][3] - object['bbox'][1]
ob_area = ob_w * ob_h
w_rate = ob_w / im_w
h_rate = ob_h / im_h
area_rate = ob_area / im_area
if not object['name'] in obs_shape.keys():
obs_shape[object['name']] = ([[ob_w,
ob_h,
ob_area,
w_rate,
h_rate,
area_rate]])
else:
obs_shape[object['name']].append([ob_w,
ob_h,
ob_area,
w_rate,
h_rate,
area_rate])
if object['name'] not in classnames:
classnames.append(object['name'])