划分模版
百度网盘链接:https://pan.baidu.com/s/11hop3cJhdri_QSQJc19ReQ
提取码:cnif
代码如下
import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join
import random
from shutil import copyfile
classes = ["one_day"]
# classes=["ball",“people”]
TRAIN_RATIO = 80 # 训练集占 10分之8
VALID_RATIO = 10 # 验证集占 10分之1
TEST_RATIO = 10 # 测试集占 10分之1
def clear_hidden_files(path): # 清理隐藏的文件
dir_list = os.listdir(path) # 以列表的形式获取目录中的文件
for i in dir_list:
abspath = os.path.join(os.path.abspath(path), i) # 获取当前文件的绝对位置并与文件拼接
if os.path.isfile(abspath): # 判断是否为一个文件
if i.startswith("._"): # 判断字符串是否以指定字符串开始
os.remove(abspath) # 删除文件
else:
clear_hidden_files(abspath) # 闭环执行,直到目录中所有的指定文件删除
def convert(size, box):
dw = 1. / size[0]
dh = 1. / size[1]
x = (box[0] + box[1]) / 2.0
y = (box[2] + box[3]) / 2.0
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(image_id): # 转换xml文件
in_file = open('datasets/Sum/Annotations/%s.xml' % image_id) # 打开xml文件的目录
out_file = open('datasets/Sum/YOLOLabels/%s.txt' % image_id, 'w') # 创建一个存放txt文件的目录
tree = ET.parse(in_file) # 载入数据
root = tree.getroot() # 获取根节点
size = root.find('size')
w = int(size.find('width').text)
h = int(size.find('height').text)
for obj in root.iter('object'): # 迭代项目
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')
in_file.close()
out_file.close()
wd = os.getcwd() # 获取当前python文件的目录
print(wd)
data_base_dir = os.path.join(wd, "datasets/") # 拼接文件目录
if not os.path.isdir(data_base_dir): # 判断是否为一个文件目录
os.mkdir(data_base_dir) # 否则就创建一个目录
work_sapce_dir = os.path.join(data_base_dir, "Sum/") # 拼接文件目录
if not os.path.isdir(work_sapce_dir):
os.mkdir(work_sapce_dir)
annotation_dir = os.path.join(work_sapce_dir, "Annotations/") # 拼接文件目录
if not os.path.isdir(annotation_dir):
os.mkdir(annotation_dir)
clear_hidden_files(annotation_dir) # 清理隐藏的文件
image_dir = os.path.join(work_sapce_dir, "JPGImages/") # 拼接文件目录
if not os.path.isdir(image_dir):
os.mkdir(image_dir)
clear_hidden_files(image_dir) # 清理隐藏的文件
yolo_labels_dir = os.path.join(work_sapce_dir, "YOLOLabels/") # 拼接文件目录
if not os.path.isdir(yolo_labels_dir):
os.mkdir(yolo_labels_dir)
clear_hidden_files(yolo_labels_dir) # 清理隐藏的文件
yolov8_train_dir = os.path.join(data_base_dir, "train/") # 训练集文件夹 1
if not os.path.isdir(yolov8_train_dir):
os.mkdir(yolov8_train_dir)
clear_hidden_files(yolov8_train_dir)
yolov8_valid_dir = os.path.join(data_base_dir, "valid/") # 拼接文件目录 -1
if not os.path.isdir(yolov8_valid_dir):
os.mkdir(yolov8_valid_dir)
clear_hidden_files(yolov8_valid_dir)
yolov8_test_dir = os.path.join(data_base_dir, "test/")
if not os.path.isdir(yolov8_test_dir):
os.mkdir(yolov8_test_dir)
clear_hidden_files(yolov8_test_dir)
yolov8_train_images_dir = os.path.join(yolov8_train_dir, "images/") # 训练集 2
if not os.path.isdir(yolov8_train_images_dir):
os.mkdir(yolov8_train_images_dir)
for filename in os.listdir(yolov8_train_images_dir):
file_path = os.path.join(yolov8_train_images_dir, filename)
if os.path.isfile(file_path):
os.remove(file_path)
clear_hidden_files(yolov8_train_images_dir)
yolov8_train_labels_dir = os.path.join(yolov8_train_dir, "labels/") # 训练集 3
if not os.path.isdir(yolov8_train_labels_dir):
os.mkdir(yolov8_train_labels_dir)
for filename in os.listdir(yolov8_train_labels_dir):
file_path = os.path.join(yolov8_train_labels_dir, filename)
if os.path.isfile(file_path):
os.remove(file_path)
clear_hidden_files(yolov8_train_labels_dir)
yolov8_valid_images_dir = os.path.join(yolov8_valid_dir, "images/") # 验证集 -2
if not os.path.isdir(yolov8_valid_images_dir):
os.mkdir(yolov8_valid_images_dir)
for filename in os.listdir(yolov8_valid_images_dir):
file_path = os.path.join(yolov8_valid_images_dir, filename)
if os.path.isfile(file_path):
os.remove(file_path)
clear_hidden_files(yolov8_valid_images_dir)
yolov8_valid_labels_dir = os.path.join(yolov8_valid_dir, "labels/") # 验证集 -3
if not os.path.isdir(yolov8_valid_labels_dir):
os.mkdir(yolov8_valid_labels_dir)
for filename in os.listdir(yolov8_valid_labels_dir):
file_path = os.path.join(yolov8_valid_labels_dir, filename)
if os.path.isfile(file_path):
os.remove(file_path)
clear_hidden_files(yolov8_valid_labels_dir)
yolov8_test_images_dir = os.path.join(yolov8_test_dir, "images/")
if not os.path.isdir(yolov8_test_images_dir):
os.mkdir(yolov8_test_images_dir)
for filename in os.listdir(yolov8_test_images_dir):
file_path = os.path.join(yolov8_test_images_dir, filename)
if os.path.isfile(file_path):
os.remove(file_path)
clear_hidden_files(yolov8_test_images_dir)
yolov8_test_labels_dir = os.path.join(yolov8_test_dir, "labels/")
if not os.path.isdir(yolov8_test_labels_dir):
os.mkdir(yolov8_test_labels_dir)
for filename in os.listdir(yolov8_test_labels_dir):
file_path = os.path.join(yolov8_test_labels_dir, filename)
if os.path.isfile(file_path):
os.remove(file_path)
clear_hidden_files(yolov8_test_labels_dir)
train_file = open(os.path.join(wd, "yolov8_train.txt"), 'w') # 创建一个拼接好的训练集路径日记
val_file = open(os.path.join(wd, "yolov8_valid.txt"), 'w') # 创建一个拼接好的验证集路径日记
test_file = open(os.path.join(wd, "yolov8_test.txt"), 'w')
train_file.close()
val_file.close()
test_file.close()
train_file = open(os.path.join(wd, "yolov8_train.txt"), 'a') # 使文件可追加
val_file = open(os.path.join(wd, "yolov8_valid.txt"), 'a') # 使文件可追加
test_file = open(os.path.join(wd, "yolov8_test.txt"), 'a')
list_imgs = os.listdir(image_dir) # 以列表的形式获取目录中的文件
prob = random.randint(1, 100)
print("Probability: %d" % prob)
for i in range(0, len(list_imgs)):
path = os.path.join(image_dir, list_imgs[i]) # 指定每一个图片的路径
if os.path.isfile(path): # 判断是否为一个文件
image_path = image_dir + list_imgs[i] # 每一个图片的绝对路径名
voc_path = list_imgs[i] # 每一个图片路径名
(nameWithoutExtention, extention) = os.path.splitext(os.path.basename(image_path)) # 返回每一个图片路径最后一级的文件名与后缀组成的元组
(voc_nameWithoutExtention, voc_extention) = os.path.splitext(os.path.basename(voc_path))
annotation_name = nameWithoutExtention + '.xml' # 每一个xml文件的绝对路径名
annotation_path = os.path.join(annotation_dir, annotation_name) # 指定Annotations中每一个xml文件
label_name = nameWithoutExtention + '.txt'
label_path = os.path.join(yolo_labels_dir, label_name) # 指定YOLOLabels中的每一个txt文件
prob = random.randint(1, 100)
print("Probability: %d" % prob)
if (prob < TRAIN_RATIO): # 训练集
if os.path.exists(annotation_path): # 判断一个路径是否存在
train_file.write(image_path + '\n') # 把图片路径写进日记中
convert_annotation(nameWithoutExtention) # 转换数据
copyfile(image_path, yolov8_train_images_dir + voc_path) # 把JPGImages中对应的图片复制到images_train文件夹中
copyfile(label_path, yolov8_train_labels_dir + label_name) # 把YOLOLabels中对应的txt文件复制到labels_train文件夹中
elif (prob > TRAIN_RATIO and prob < TRAIN_RATIO+VALID_RATIO): # 验证集
if os.path.exists(annotation_path):
val_file.write(image_path + '\n')
convert_annotation(nameWithoutExtention) # 转换数据
copyfile(image_path, yolov8_valid_images_dir + voc_path)
copyfile(label_path, yolov8_valid_labels_dir + label_name)
else:
if os.path.exists(annotation_path):
test_file.write(image_path + '\n')
convert_annotation(nameWithoutExtention)
copyfile(image_path, yolov8_test_images_dir + voc_path)
copyfile(label_path, yolov8_test_labels_dir + label_name)
train_file.close()
val_file.close()
test_file.close()