yolov4数据集制作与规范性检查

1、图片重命名->rename.py

import os
#import cv2
import shutil

jpg_path = '/media/em/data_1/0716/data2/socket/mark'
xml_path = '/media/em/data_1/0716/data2/socket/xml/'

out_jpg_path = '/media/em/data_1/0716/data2/socket/JPEGImages1/'
out_xml_path = '/media/em/data_1/0716/data2/socket/Annotations1/'

if not os.path.exists(out_jpg_path):
    os.mkdir(out_jpg_path)
    print(out_jpg_path + '----创建成功')

if not os.path.exists(out_xml_path):
    os.mkdir(out_xml_path)
    print(out_xml_path + '----创建成功')

list_jpg = os.listdir(jpg_path)
#list_xml =os.listdir(xml_path)

#i = len(list_jpg)
i = 20000
for list_jpgs in list_jpg:
    total_num_fil = len(list_jpg)
    #no_houzhui = list_jpgs.split('.')[0]
    no_houzhui = os.path.splitext(list_jpgs)[0]
    print(no_houzhui)
    #print(tmp[0])
    jpg_name = jpg_path + '/' + str(no_houzhui) + '.jpg'
    xml_name = xml_path + str(no_houzhui) + '.xml'
    dst_jpg = out_jpg_path + str(i) + '.jpg'
    shutil.copy(jpg_name, dst_jpg)
    #os.rename(jpg_name,dst_jpg)

    dst_xml = out_xml_path + str(i) + '.xml'
    print(jpg_name)
    shutil.copy(xml_name, dst_xml)
    #os.rename(xml_name,dst_xml)
    i=i+1
    print(i)

print("finish!")

2、数据规范性检查->xml_jpg

# -*- coding: utf-8 -*-
#根据单张图片的xml文件，在图片上标注出目标
import xml.etree.ElementTree as ET
import os
import cv2

 

jpg_path = '/media/em/data_1/0716/data2/socket/JPEGImages1'
xml_path = '/media/em/data_1/0716/data2/socket/Annotations1'

list_jpg = os.listdir(jpg_path)
list_xml = os.listdir(xml_path)

color = (4, 250, 7)

for list_jpgs in list_jpg:
    print('++++++++++++++++++++++++++++++++++++++++++++++++++++++++++')
    print(list_jpgs)
    no_houzhui = os.path.splitext(list_jpgs)[0]
    xml_file = xml_path + '/' + no_houzhui + '.xml'
    tree = ET.parse(xml_file)
    root = tree.getroot()
    jpg_file = jpg_path + '/' + no_houzhui +'.jpg'
    im = cv2.imread(jpg_file)
    print(im.shape)
    w = im.shape[0]
    h = im.shape[1]
    for object in root.findall('object'):
        object_name = object.find('name').text
        Xmin = int(object.find('bndbox').find('xmin').text)
        Ymin = int(object.find('bndbox').find('ymin').text)
        Xmax = int(object.find('bndbox').find('xmax').text)
        Ymax = int(object.find('bndbox').find('ymax').text)
        if(Xmin < 0 or Xmax >= h or Ymin < 0 or Ymax >= w or Xmin >= Xmax or Ymin >= Ymax):
            print('Xmin---Xmax---Ymin---Ymax')
            print(Xmin,Xmax,Ymin,Ymax)
            print("data error！！！！")
            cv2.rectangle(im, (Xmin, Ymin), (Xmax, Ymax), color, 2)
            font = cv2.FONT_HERSHEY_SIMPLEX
            cv2.putText(im, object_name, (Xmin, Ymin - 7), font, 0.5, (6, 230, 230), 2)
            dst = cv2.resize(im, (640, 480))
            cv2.imshow('01', dst)
            cv2.waitKey(0)
            continue
        else:
            cv2.rectangle(im, (Xmin, Ymin), (Xmax, Ymax), color, 2)
            font = cv2.FONT_HERSHEY_SIMPLEX
            cv2.putText(im, object_name, (Xmin, Ymin - 7), font, 0.5, (6, 230, 230), 2)
            dst = cv2.resize(im, (640, 480))
            cv2.imshow('01', dst)
            #cv2.waitKey()
    cv2.waitKey(0)

3、yolo数据集划分与label生成

3.1、train.txt，trainval.txt，test.txt，val.txt生成

import os
import random
import sys

#if len(sys.argv) < 2:
#    print("no directory specified, please input target directory")
#    exit()

#root_path = sys.argv[1]
root_path = '/media/em/data_1/0716/VOCdevkit/VOC2007'

xmlfilepath = root_path + '/Annotations'

txtsavepath = root_path + '/ImageSets/Main'

if not os.path.exists(root_path):
    print("cannot find such directory: " + root_path)
    exit()

if not os.path.exists(txtsavepath):
    os.makedirs(txtsavepath)

trainval_percent = 0.9
train_percent = 0.8
total_xml = os.listdir(xmlfilepath)
num = len(total_xml)
list = range(num)
tv = int(num * trainval_percent)
tr = int(tv * train_percent)
trainval = random.sample(list, tv)
train = random.sample(trainval, tr)

print("train and val size:", tv)
print("train size:", tr)

ftrainval = open(txtsavepath + '/trainval.txt', 'w')
ftest = open(txtsavepath + '/test.txt', 'w')
ftrain = open(txtsavepath + '/train.txt', 'w')
fval = open(txtsavepath + '/val.txt', 'w')

for i in list:
    name = total_xml[i][:-4] + '\n'
    if i in trainval:
        ftrainval.write(name)
        if i in train:
            ftrain.write(name)
        else:
            fval.write(name)
    else:
        ftest.write(name)

ftrainval.close()
ftrain.close()
fval.close()
ftest.close()
3.2、yolov4的label和数据集划分

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join

sets=[('2007', 'train'), ('2007', 'val'), ('2007', 'test')]

#可根据自己类别修改

classes = ["background_waiguan_obj", "chetou", "chewei",
           "chepai", "chebiao", "qianzhudeng", "qianfudeng",
           "houzhudeng", "jiaotaban", "xinglijia", "lungu", "tiehua",
           "fanghuzhuangzhi", "weiyi", "toukui_mtc", "paiqikong", "cheshen",
           "houshijing", "NUM_T", "socket",  "luntai", "chelian","sanjiaojia", "fanguangbeixin"]

def convert(size, box):
    dw = 1./(size[0])
    dh = 1./(size[1])
    x = (box[0] + box[1])/2.0 - 1
    y = (box[2] + box[3])/2.0 - 1
    w = box[1] - box[0]
    h = box[3] - box[2]
    x = x*dw
    w = w*dw
    y = y*dh
    h = h*dh
    return (x,y,w,h)

def convert_annotation(year, image_id):
    in_file = open('/media/em/data_1/0716/VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id))
    out_file = open('/media/em/data_1/0716/VOCdevkit/VOC%s/labels/%s.txt'%(year, image_id), 'w')
    tree=ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    w = int(size.find('width').text)
    h = int(size.find('height').text)

    for obj in root.iter('object'):
        difficult = obj.find('difficult').text
        cls = obj.find('name').text
        if cls not in classes or int(difficult)==1:
            continue
        cls_id = classes.index(cls)
        xmlbox = obj.find('bndbox')
        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text), float(xmlbox.find('ymax').text))
        bb = convert((w,h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')

#wd = getcwd()

for year, image_set in sets:
    if not os.path.exists('/media/em/data_1/0716/VOCdevkit/VOC%s/labels/'%(year)):
        os.makedirs('/media/em/data_1/0716/VOCdevkit/VOC%s/labels/'%(year))
    image_ids = open('/media/em/data_1/0716/VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split()
    list_file = open('/media/em/data_1/0716/VOCdevkit/%s_%s.txt'%(year, image_set), 'w')
    for image_id in image_ids:
        list_file.write('/media/em/data_1/0716/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%( year, image_id))
        convert_annotation(year, image_id)
    list_file.close()
print('finish!')

#os.system("cat 2007_train.txt 2007_test.txt > train.txt")
#os.system("cat 2007_train.txt 2007_val.txt 2007_test.txt > train.all.txt")

至此yolov4生成2007_train.txt，2007_test.txt，2007_val.txt和labels(文件夹)，数据集制作完成