文本检测训练结果评测

1. 评测链接地址参照：http://rrc.cvc.uab.es/?ch=4&com=mymethods&task=1

2. 将测试的输出结果压缩为zip包，代码如下：

import os
import xml.etree.ElementTree as ET
import numpy as np
import zipfile
from functools import reduce

#src_dir = '/home/hhh/data/labelData_6w/ImageSets/lit.txt'
src_dir = '/home/hhh/data/labelData_6w/ImageSets/test_complex.txt'
eval_dir = "/home/hhh/code/EAST/output_complex"
dst_dir = "/home/hhh/data/labelData_6w/Annotations_det/complex"
# generate zip file
eval_output='/home/hhh/complex_submit.zip'
gt_output = '/home/hhh/complex_gt.zip'

def polygon_area(poly):
'''
compute area of a polygon
:param poly:
:return:
'''
edge = [
(poly[1][0] - poly[0][0]) * (poly[1][1] + poly[0][1]),
(poly[2][0] - poly[1][0]) * (poly[2][1] + poly[1][1]),
(poly[3][0] - poly[2][0]) * (poly[3][1] + poly[2][1]),
(poly[0][0] - poly[3][0]) * (poly[0][1] + poly[3][1])
]
return np.sum(edge)/2.

if __name__ == "__main__":
if not os.path.exists(dst_dir):
os.makedirs(dst_dir)
with open(src_dir, 'r') as f:
xml_list = [os.path.join('/home/data/vocData/labelData_6w/Annotations', ff.strip() + '.xml') for ff in f.readlines()]
for xmlfile in xml_list:
pre, back = os.path.splitext(os.path.basename(xmlfile))
tree = ET.parse(xmlfile)
root = tree.getroot()

with open(os.path.join(dst_dir, pre+".txt"), "w") as f:
for obj in root.iter('object'):
box = obj.find('bndbox')
x1 = int(box.find('x1').text)
y1 = int(box.find('y1').text)
x4 = int(box.find('x2').text)
y4 = int(box.find('y2').text)
x3 = int(box.find('x3').text)
y3 = int(box.find('y3').text)
x2 = int(box.find('x4').text)
y2 = int(box.find('y4').text)
label = obj.find('class').text
if label is None:
label = '###'
if '\n' in label:
label = reduce(lambda x,y:x+y, label.split('\n'))
text_poly = np.array([[x1, y1], [x2, y2], [x3, y3], [x4, y4]])
area = polygon_area(text_poly)
if area > 0:
print('gt poly in wrong direction')
poly = text_poly[(0, 3, 2, 1), :]
(x1, y1), (x2, y2), (x3, y3), (x4, y4) = poly.tolist()
line = str(x1)+','+str(y1)+','+str(x2)+','+str(y2)+','+str(x3)+','+str(y3)+','+str(x4)+','+str(y4)+','+ label +'\n'

f.write(line)

# zip submit file
arch = zipfile.ZipFile(eval_output,'w')
files = [os.path.join(eval_dir,f) for f in os.listdir(eval_dir) if f.endswith('.txt')]
for txtfile in files:
with open(txtfile, 'r') as f:
s=''
for line in f.readlines():
x1, y1, x2, y2, x3, y3, x4, y4 = list(map(int,line.strip().split(',')))
text_poly = np.array([[x1, y1], [x2, y2], [x3, y3], [x4, y4]])
area = polygon_area(text_poly)
if area > 0:
print('submit poly in wrong direction')
poly = text_poly[(0, 3, 2, 1), :]
(x1, y1), (x2, y2), (x3, y3), (x4, y4) = poly.tolist()
newline = str(x1)+','+str(y1)+','+str(x2)+','+str(y2)+','+str(x3)+','+str(y3)+','+str(x4)+','+str(y4)+'\n'
s += newline
arch.writestr(os.path.basename(txtfile), s)

# zip gt file
arch2 = zipfile.ZipFile(gt_output,'w')
names = [f.split('/')[-1] for f in files]
for text in [os.path.join(dst_dir,f) for f in os.listdir(dst_dir) if f.endswith('.txt')]:
if text.split('/')[-1] in names:
with open(text, 'r') as f:
s=''
for line in f.readlines():
x1, y1, x2, y2, x3, y3, x4, y4 = list(map(int,line.strip().split(',')[0:8]))
if len(line.split(',')) == 9:
label = line.strip().split(',')[-1]
else:
lable = reduce(lambda x,y:x+y,line.strip().split(',')[8:])
newline = str(x1)+','+str(y1)+','+str(x2)+','+str(y2)+','+str(x3)+','+str(y3)+','+str(x4)+','+str(y4)+',' + label +'\n'
s += newline
arch2.writestr(os.path.basename(text),s)

3. 将压缩包转换测试所需要的格式，代码如下

import zipfile
import os
import numpy as np

src_gt_path = '/home/hhh/complex_gt.zip'
src_submit_path = '/home/hhh/complex_submit.zip'
dst_gt_path = '/home/hhh/tmp/complex_gt_result.zip'
dst_submit_path = '/home/hhh/tmp/complex_submit_result.zip'

try:
archive=zipfile.ZipFile(src_gt_path, mode='r', allowZip64=True)
arc=zipfile.ZipFile(src_submit_path, mode='r', allowZip64=True)
except :
raise Exception('Error loading the ZIP archive.')

if len(archive.namelist()) != len(arc.namelist()):
raise Exception('number of file in zipfile is equal.')

print('convert gt data ...')
name_dict = {}
idx=1
zip_gt_result = zipfile.ZipFile(dst_gt_path, mode='w')
for name in archive.namelist():
f = archive.open(name, 'r')
name_dict[name] = idx
s = ''
for data in f.readlines():
#print(str(data, encoding='utf-8'))
s += str(data, encoding='utf-8')
zip_gt_result.writestr('gt_img_' + str(idx) + '.txt', s)
idx = idx + 1

print('convert submit data ...')
zip_submit_result = zipfile.ZipFile(dst_submit_path, mode='w')
for name in arc.namelist():
#print(name)
ff = arc.open(name, 'r')
s2 = ''
for data in ff.readlines():
#print(str(data, encoding='utf-8'))
s2 += str(data,encoding='utf-8')
zip_submit_result.writestr('res_img_' + str(name_dict[name]) + '.txt', s2)

4. 最后，生成完需要的压缩包格式后进行测试，执行下面代码得到结果

python script.py -g=../tmp/complex_gt_result.zip -s=../tmp/complex_submit_result.zip -o=../tmp

文本检测训练结果评测相关推荐

PaddleOCR 文本检测训练+推理模型转换教程
GitHub - PaddlePaddle/PaddleOCR at release/2.1 一.环境准备 pip3 install --upgrade pip如果您的机器安装的是CUDA9或CUDA ...
CVPR 2022 | 阿里华科提出：针对场景文本检测的视觉语言模型预训练
点击下方卡片,关注"CVer"公众号 AI/CV重磅干货,第一时间送达点击进入-> CV 微信技术交流群转载自:CSIG文档图像分析与识别专委会本文简要介绍了发表于CV ...
paddleocr文本检测模型的训练
1.环境的安装和开源项目的下载首先我个人建议,玩深度学习的话,不管是工作还是学习,最起码要配一个有GPU的电脑.我个人有着血淋淋的教训,我本人是电气工程的一名学生,本科期间一点深度学习和机器学习的基 ...
（二）目标检测模型的评测与训练技巧
转载自知乎:https://zhuanlan.zhihu.com/p/34142321 关于作者: @李家丞同济大学数学系本科在读,现为格灵深瞳算法部实习生. -------------------- ...
深度学习开源数据集——自动驾驶、目标检测、人脸识别、文本检测、图像分类
前言在深度学习中,如果没有数据集,就无法训练模型,所以数据是根本,下面列出几个常用数据集. 想要更多数据集,可以去这个地址:https://www.cvmart.net/dataSets或https ...
格物钛数据平台国内外经典开源数据汇总（自动驾驶、目标检测、人脸识别、人体姿态估计、文本检测、NLP、医疗）
本文整理了国内外经典的开源数据,包含了目标检测.自动驾驶.人脸识别.自然语言处理.文本检测.医疗等方向,具体如下. 一.自动驾驶领域数据集 KITTI数据集 KITTI数据集由德国卡尔斯鲁厄理工学院和 ...
连通域最小外接矩形算法原理_基于分割的文本检测算法之PSENet/PAN/DBNet
1. 文本检测难点文本内包含文本,艺术字体,任意方向 ,曲线文字 ,多语言,其他环境因素等是文本检测中的难点 2. 分割问题1: 语义分割模型是对pixel进行分类,所以理论上讲,可以检测不规则的 ...
mfc倾斜文本输入_文本检测知识梳理（持续更新）
最近在做作业批改场景的OCR相关算法研发工作,打算梳理一下文本检测的相关知识,也欢迎大家留言讨论. 目前主流的基于深度学习的目标检测方法大体分为两类:one-stage和two-stage: 1.Tw ...
人工智能学习--文本检测实践
注释:文本检测和文本识别是两回事. 可能现在已经有 end-to-end的深度神经网络可以将文本检测和识别一起实现,这个要去搜相关的sci论文. 文本检测,是从一张图片中找到文字区域,并用矩形框标 ...

文本检测训练结果评测

文本检测训练结果评测相关推荐

最新文章

热门文章