yolov5 win10 gpu 初探目标识别过滑块验证码

1，环境安装

不懂得自己百度一下找教程。不在叙述。下面是笔记
环境。WIN10LTS，GTX750TI

建议不好用CPU，时间太漫长。翻箱底翻了个

conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo/
conda config --set show_channel_urls yes

conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --set show_channel_urls yes

conda create -n yolov5 python==3.8
询问是否下载，输入y
activate yolov5
cd yolov5
查看cuda版本，下载相应版本pytroch和cudnn
import torch
torch.cuda.is_available()

conda env remove -n yolov5_cuda10.2

验证CUDA是否可用
import torch
torch.cuda.is_available()

conda env remove -n yolov5_cuda10.2

版本必须对于上。离线本地安装
pip install torch-2.1.1+cu121-cp38-cp38-win_amd64.whl
pip install torchvision-0.16.1+cpu-cp38-cp38-win_amd64.whl
pip install torchvision==0.16.1 –index-url https://download.pytorch.org/whl/cu121

运行例子缺少组件加这个
import os
os.environ[“GIT_PYTHON_REFRESH”] = “quiet”

使用labelimg标记图片

导出为YOLO得格式。我这里用了初略200个试了下效果可以。后面增加到500个。精度很高了

训练

python train.py –weights yolov5s.pt –batch-size=4 –epochs 40 –data=./data/sf.yaml

path: C:\Users\Administrator\yolov5\data\sf
train: C:\Users\Administrator\yolov5\data\sf\train  # train images (relative to 'path') 128 images
val: C:\Users\Administrator\yolov5\data\sf\val  # val images (relative to 'path') 128 images
test: 
 
# Classes
names:
  0: 0

查看验证集结果，效果很好，都很精确

部署FLASK接口供其他程序使用

import argparse
import csv
import os
import platform
import sys
from pathlib import Path
import numpy as np
import torch

FILE = Path(__file__).resolve()
ROOT = FILE.parents[0]  # YOLOv5 root directory
if str(ROOT) not in sys.path:
    sys.path.append(str(ROOT))  # add ROOT to PATH
ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative

from ultralytics.utils.plotting import Annotator, colors, save_one_box

from models.common import DetectMultiBackend
from utils.dataloaders import IMG_FORMATS, VID_FORMATS, LoadImages, LoadScreenshots, LoadStreams
from utils.general import (LOGGER, Profile, check_file, check_img_size, check_imshow, check_requirements, colorstr, cv2,
                           increment_path, non_max_suppression, print_args, scale_boxes, strip_optimizer, xyxy2xywh)
from utils.torch_utils import select_device, smart_inference_mode

from models.experimental import attempt_load
import base64

from flask import Flask, request, Response,render_template
import json
import cv2
import time

from PIL import Image
app = Flask(__name__)



UPLOAD_FOLDER = r'./uploads'
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
def base64_to_image(base64_code): 
    # base64解码
    img_data = base64.b64decode(base64_code)
    # 转换为np数组
    img_array = np.fromstring(img_data, np.uint8)
    # 转换成opencv可用格式
    image_base64_dec = cv2.imdecode(img_array, cv2.COLOR_RGB2BGR)
    return image_base64_dec
    
    
def redirect(source):
    #source = 'C:/Users/Administrator/yolov5/data/images/1704215184961.bmp',
    stride = int(model.stride.max())
    imgsz = check_img_size(640, s=stride)
    dataset = LoadImages(source)
    #seen, windows, dt = 0, [], (Profile(), Profile(), Profile())
    for path, im, im0s, vid_cap, s in dataset:
        
        im = torch.from_numpy(im).to(device)
        im = im.half() if model.half else im.float()  # uint8 to fp16/32
        im /= 255  # 0 - 255 to 0.0 - 1.0
        if len(im.shape) == 3:
            im = im[None]  # expand for batch dim
        pred = model(im, augment=False, visualize=False)
        pred = non_max_suppression(pred, conf_thres, iou_thres, None, False, max_det=1000)
        seen = 0
        for i, det in enumerate(pred):  # per image
            seen += 1
            p, im0, frame = path, im0s.copy(), getattr(dataset, 'frame', 0)
            if len(det):
                det[:, :4] = scale_boxes(im.shape[2:], det[:, :4], im0.shape).round()
                for *xyxy, conf, cls in reversed(det):
                    
                    c1, c2 = (int(xyxy[0]), int(xyxy[1])), (int(xyxy[2]), int(xyxy[3]))
                    #print("左上点的坐标为：(" + str(c1[0]) + "," + str(c1[1]) + ")，右下点的坐标为(" + str(c2[0]) + "," + str(c2[1]) + ")")
                    print("识别坐标为：" +str(c1[0]))
                    return str(c1[0])
        return



def allowed_file(filename):
    return '.' in filename and filename.rsplit('.', 1)[1] in ALLOWED_EXTENSIONS
 
 

@app.route('/detect', methods=['POST'])
def detect():
    file = request.files['file']
    
    src_path = os.path.join(app.config['UPLOAD_FOLDER'], file.filename)
    file.save(src_path)
    #source = src_path
    print(src_path)
    results = redirect(src_path)
    return results
    # do something with the results
    #return jsonify({'result': results})
if __name__ == "__main__":
    #opt=parse_opt()
    #run(**vars(opt))
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    model = attempt_load(ROOT / 'yolov5s.pt')
    model.to(device).eval()
    model.half()
    conf_thres = 0.65  # NMS置信度
    iou_thres = 0.45  # IOU阈值

    #print(opt)
    app.run(host='0.0.0.0', port=5001, debug=True)

最终效果

返回缺口坐标，很好，收工。

一些参考连接

https://blog.csdn.net/m0_58892312/article/details/120923608
https://blog.csdn.net/ml1211449109/article/details/119811567
https://blog.csdn.net/weixin_52797278/article/details/125427898
https://zhuanlan.zhihu.com/p/513967538
https://blog.csdn.net/wangzhuanjia/article/details/125278833