yolov4实战训练数据

1、克隆项目文件

项目Github地址：https://github.com/AlexeyAB/darknet
打开终端，克隆项目

git clone https://github.com/AlexeyAB/darknet.git

无法克隆的话，把https修改为git

git clone git://github.com/AlexeyAB/darknet.git

修改Makefile文件

GPU=1
CUDNN=1
OPENCV=1

编译项目

cd darknet
make 或者 make -j8（加速编译）

2、下载训练的权重文件

链接：(tiny) YOLOv4 详细训练指南(附下载链接)

3、准备数据集

可以用labelimg标注数据，然后图片放image、标注文件放xml

4、修改.data、.name、.cfg文件

根据自己训练数据的类别修改.data、.name文件
.cfg文件注意修改classes=你的类别数，并修改classes往上找的第一个filters=（classes+5）x 3

5、生成训练集和验证集（split_train_val.py）

# coding:utf-8
import os
import random
import argparseparser = argparse.ArgumentParser()
#xml文件的地址，根据自己的数据进行修改 xml一般存放在Annotations下
parser.add_argument('--xml_path', default='Fincir\\xml', type=str, help='input xml label path')
#数据集的划分，地址选择自己数据下的ImageSets/Main，没有的话新建一个
parser.add_argument('--txt_path', default='jht_data\\main', type=str, help='output txt label path')
opt = parser.parse_args()trainval_percent = 1.0
train_percent = 0.9
xmlfilepath = opt.xml_path
txtsavepath = opt.txt_path
total_xml = os.listdir(xmlfilepath)
if not os.path.exists(txtsavepath):os.makedirs(txtsavepath)num = len(total_xml)
list_index = range(num)
tv = int(num * trainval_percent)
tr = int(tv * train_percent)
trainval = random.sample(list_index, tv)
train = random.sample(trainval, tr)file_trainval = open(txtsavepath + '/trainval.txt', 'w')
file_test = open(txtsavepath + '/test.txt', 'w')
file_train = open(txtsavepath + '/train.txt', 'w')
file_val = open(txtsavepath + '/val.txt', 'w')for i in list_index:name = total_xml[i][:-4] + '\n'if i in trainval:file_trainval.write(name)if i in train:file_train.write(name)else:file_val.write(name)else:file_test.write(name)file_trainval.close()
file_train.close()
file_val.close()
file_test.close()

6、xml转txt文件，生成图片、txt文件路径（voc_label.py）

# -*- coding: utf-8 -*-
import xml.etree.ElementTree as ET
import os
from os import getcwdsets = ['train', 'val', 'test']
classes = ["0", "1", "2"]   # 改成自己的类别
# classes = ["0", "1", "2", "3", "4", "5", "6", "7", "8", "9","A","B","C","D","E","F","G","H","I","J","K","L","M","N",
#            "O","P","Q","R","S","T","U","V","W","X","Y","Z","-","/"]   # 改成自己的类别
abs_path = os.getcwd()
print(abs_path)def convert(size, box):dw = 1. / (size[0])dh = 1. / (size[1])x = (box[0] + box[1]) / 2.0 - 1y = (box[2] + box[3]) / 2.0 - 1w = box[1] - box[0]h = box[3] - box[2]x = x * dww = w * dwy = y * dhh = h * dhreturn x, y, w, hdef convert_annotation(image_id):in_file = open('Fincir/xml/%s.xml' % (image_id), encoding='UTF-8')out_file = open('Fincir/image/%s.txt' % (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#difficult = obj.find('Difficult').textcls = obj.find('name').textif cls not in classes or int(difficult) == 1:continuecls_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))b1, b2, b3, b4 = b# 标注越界修正if b2 > w:b2 = wif b4 > h:b4 = hb = (b1, b2, b3, b4)bb = convert((w, h), b)out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')wd = getcwd()
for image_set in sets:image_ids = open('jht_data\\xml\\%s.txt' % (image_set)).read().strip().split()list_file = open('%s.txt' % (image_set), 'w')for image_id in image_ids:list_file.write(abs_path + '\\Fincir\\image\\%s.bmp\n' % (image_id))convert_annotation(image_id)list_file.close()

7、训练

//linux
./darknet detector train build/darknet/x64/cir_data/voc_cir.data 
build/darknet/x64/cir_data/yolov4-tiny.cfg 
build/darknet/x64/cir_data/yolov4-tiny.conv.29 -gpus -map
./darknet detector train train_data/voc.data train_data/yolov4.cfg train_data/yolov4.conv.137 -gpus -map ./darknet detector train train_data/voc.data train_data/yolov4-tiny.cfg train_data/yolov4-tiny.conv.29 -gpus -map

8、验证权重文件

./darknet detector map train_data/voc.data train_data/yolov4-tiny.cfg backup/yolov4-tiny_best.weights