Samples
Overview
LogoDet-3K是一个用于检测logo的数据集,共含有大类9种,小类3000种,图片158652张,标签约200000个。下图是具体的分类情况
Data Explore
9个文件夹,每个文件里有对应的子类文件夹,每个子类文件夹里是图片和对应的xml格式标签,xml标签格式如下
<?xml version="1.0"?>
<annotation verified="no">
<folder>VOC2007</folder>
<filename>1.jpg</filename>
<source>
<database>Unknown</database>
</source>
<size>
<width>484</width>
<height>383</height>
<depth>3</depth>
</size>
<segmented>0</segmented>
<object>
<name>2xist</name>
<pose>Unspecified</pose>
<truncated>0</truncated>
<difficult>0</difficult>
<bndbox>
<xmin>276</xmin>
<ymin>143</ymin>
<xmax>459</xmax>
<ymax>184</ymax>
</bndbox>
</object>
</annotation>
数据初始化
输入图片路径img_path,即可获得图片的box2d标签,其格式为:[[xmin,ymin,xmax,ymax,category],[],...]
def acquire_label_xml(img_path):
DOMTree = xml.dom.minidom.parse(img_path)
collection = DOMTree.documentElement
boundingbox = collection.getElementsByTagName("object")
img_lable = []
for i in boundingbox:
tmp = []
category = i.getElementsByTagName("name")[0].childNodes[0].data
tmp.append(float(
[j.childNodes[0].data for j in i.getElementsByTagName("bndbox")[0].getElementsByTagName("xmin")][0]))
tmp.append(float(
[j.childNodes[0].data for j in i.getElementsByTagName("bndbox")[0].getElementsByTagName("ymin")][0]))
tmp.append(float(
[j.childNodes[0].data for j in i.getElementsByTagName("bndbox")[0].getElementsByTagName("xmax")][0]))
tmp.append(float(
[j.childNodes[0].data for j in i.getElementsByTagName("bndbox")[0].getElementsByTagName("ymax")][0]))
tmp.append(category)
img_lable.append(tmp)
return img_lable
关注公众号,后台回复 LogoDet 即可获得数据集