devbox
Gausst松鼠会
这个屌丝很懒，什么也没留下！
关注作者
热门标签
热门文章
当前位置:   article > 正文

YOLOV5 训练自定义数据_yolov5怎么修改成自己上传图片地址

作者：Gausst松鼠会 | 2024-06-03 23:58:48

赞

踩

yolov5怎么修改成自己上传图片地址

YOLOV5 模型构建

  1. 部署项目

    git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

    • 1
    • 2
    • 3

  2. 准备数据

    1. 数据存为两个文件夹 images, annotation. images 存图片数据，annotation存xml文件。txt文件夹由数据集划分代码生成。
      数据文件结构为:
      data–
      ++++ images
      ++++ annotation
      ++++ txt

    2. 修改配置文件。

      # config.json
{
    "root": "data",
    "annotation": "data/annotation",
    "image": "data/images",
    "unicode": ".xml",
    "txt": "data/txt",
    "classes": ["类名1"，"类名2"，...]
}

      • 1
      • 2
      • 3
      • 4
      • 5
      • 6
      • 7
      • 8
      • 9

    以下为数据集划分脚本，train,val表示划分训练集和验证集比例。测试集占比为1-train-val;
    在这里插入代码片

    def split_data(train=0.8, val=0.1):
    # read file path from json file
    with open('cleandata/config.json', 'r') as f:
        data = json.load(f)
    # data = json.load('/home/s4/Desktop/yolo/yolov5/yolov5/cleandata/config.json')
    xml_file_path = data['annotation']
    txt_save_path = data['txt']
    img_file_path = data['image']
    total_xml = os.listdir(xml_file_path)
    total_img = os.listdir(img_file_path)
    print(txt_save_path)
    if not os.path.exists(txt_save_path):
        print(txt_save_path)
        os.makedirs(txt_save_path)
    num = len(total_xml)
    list_index = range(num)
    num_train = int(num * train)
    num_val = int(num * val)

    # random split
    train_val = random.sample(list_index, num_train + num_val)
    train = random.sample(train_val, num_train)

    # load in file
    file_train_val = open(os.path.join(txt_save_path, 'train_val.txt'), 'w')
    file_test = open(os.path.join(txt_save_path, 'test.txt'), 'w')
    file_train = open(os.path.join(txt_save_path, 'train.txt'), 'w')
    file_val = open(os.path.join(txt_save_path, 'val.txt'), 'w')

    for i in list_index:
        full_path = os.path.join(img_file_path, total_img[i])+'\n'
        file = None
        if i in train_val:
            file_train_val.write(full_path)
            if i in train:
                file = file_train
            else:
                file = file_val
            pass
        else:
            file = file_test
            pass
        file.write(full_path)
    file_train_val.close()
    file_train.close()
    file_val.close()
    file_test.close()

    pass
split_data()

    • 1
    • 2
    • 3
    • 4
    • 5
    • 6
    • 7
    • 8
    • 9
    • 10
    • 11
    • 12
    • 13
    • 14
    • 15
    • 16
    • 17
    • 18
    • 19
    • 20
    • 21
    • 22
    • 23
    • 24
    • 25
    • 26
    • 27
    • 28
    • 29
    • 30
    • 31
    • 32
    • 33
    • 34
    • 35
    • 36
    • 37
    • 38
    • 39
    • 40
    • 41
    • 42
    • 43
    • 44
    • 45
    • 46
    • 47
    • 48
    • 49
    • 50
    1. 生成用于训练的标签文件。将annotation文件夹中的xml文件进行转换生成的包括所有txt标签文件的label文件夹，txt文件每行包括：类别，x,y,w,h五个数据。
import shutil
import random
import xml.etree.ElementTree as ET
import os
import json
from tqdm import tqdm
sets = ['train', 'val', 'test']


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(image_id, annotation_path, root_path, classes):
    in_file = open(annotation_path + '/%s.xml' % (image_id),
                   encoding='UTF-8')

    out_file = open(os.path.join(root_path, 'labels') + '/%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').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))
        b1, b2, b3, b4 = b
        # 标注越界修正
        if b2 > w:
            b2 = w
        if b4 > h:
            b4 = h
        b = (b1, b2, b3, b4)
        bb = convert((w, h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')


def generate_labels():
    with open('config.json', 'r') as f:
        data = json.load(f)
        classes = data['classes']
        root_path = data['root']
        annotation_path = data['annotation']
        image_path = data['image']
        txt_path = data['txt']

    for image_set in sets:
        if not os.path.exists(os.path.join(root_path, 'labels')):
            os.makedirs(os.path.join(root_path, 'labels'))
        image_ids = open(
            txt_path + '/%s.txt' % (
                image_set)).read().strip().split()
        for image_id in tqdm(image_ids):
            convert_annotation(os.path.basename(image_id)[:-4],annotation_path,root_path,classes)
generate_labels()         

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
  • 45
  • 46
  • 47
  • 48
  • 49
  • 50
  • 51
  • 52
  • 53
  • 54
  • 55
  • 56
  • 57
  • 58
  • 59
  • 60
  • 61
  • 62
  • 63
  • 64
  • 65
  • 66
  • 67
  • 68
  • 69
  • 70
  • 71
  • 72

  1. 训练

    1. 配置yolo5x.yaml文件。

      修改yolo5x.yaml中的nc 参数为训练数据的类别数。

    2. 配置 类别 yaml文件。

      新建yaml文件,根据训练数据修改对应参数。

      train: ./data/txt/train.txt
val: ./data/txt/val.txt
test: ./data/txt/test.txt
         
# Classes
nc: 7  # number of classes
names: ["类别1","类别2",""]  # class names

      • 1
      • 2
      • 3
      • 4
      • 5
      • 6
      • 7
      python train.py --data 自定义类别.yaml --cfg yolov5n.yaml --weights yolov5n.pt --batch-size 128
                                       yolov5s                     yolov5s            
                                       yolov5m                     yolov5m          
                                       yolov5l                     yolov5l          
                                       yolov5x                     yolov5x           

      • 1
      • 2
      • 3
      • 4
      • 5

  2. 推理

    python detect.py --weights best.pt  --source 0  # webcam
                          					img.jpg  # image
                         					vid.mp4  # video
                          					path/  # directory
                          					path/*.jpg  # glob
                          					'https://youtu.be/Zgi9g1ksQHc'  # YouTube
                          					'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream

    • 1
    • 2
    • 3
    • 4
    • 5
    • 6
    • 7

  3. 模型转换

    将pt模型转换为用于TensorRT框架下的.engine文件.

    python detect.py --weights best.pt  --data 自定义类别.yaml  --include 'engine' simplify

    • 1

    生成.engine文件后使用.engine进行推理

    ython detect.py --weights best.engine  --source 0  # webcam
                          						img.jpg  # image
                         						vid.mp4  # video
                          						path/  # directory
                          						path/*.jpg  # glob
                          						'https://youtu.be/Zgi9g1ksQHc'  # YouTube
                          						'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP stream

    • 1
    • 2
    • 3
    • 4
    • 5
    • 6
    • 7
声明：本文内容由网友自发贡献，不代表【wpsshop博客】立场，版权归原作者所有，本站不承担相应法律责任。如您发现有侵权的内容，请联系我们。转载请注明出处：https://www.wpsshop.cn/w/Gausst松鼠会/article/detail/669474
推荐阅读
相关标签

Copyright © 2003-2013 www.wpsshop.cn 版权所有，并保留所有权利。

  

闽ICP备14008679号