本文共 5750 字，大约阅读时间需要 19 分钟。

是一个项目的一个功能之一，调试了两小时，终于能够

javascript设置开始计和暂停计时 监控人脸 记录时间了 效果图： 离开页面之后回到页面会从0计时（不是关闭页面，而是页面失去焦点） 离开摄像头时会弹出提示。 离开摄像头反馈给后端的时间。

全部代码：

其中包含了图片解码编码，打开摄像头，获取后端结果。

服务器有两个，一个是python的pytorch深度学习处理图片 在flask框架下，一个是java的springboot来获得离开摄像头的时间。 java部分： 实体类：

package com.naughty.userlogin02.bean;import lombok.Data;@Datapublic class Nowtime {
       int id;     String timenot;   String total;  String  nowtime;  String begintime;}

跨域请求：

@RestControllerpublic class Timecontroller {
       @Autowired    TimeDao timeDao;    static int id = 0;    @CrossOrigin    @PostMapping("/gettime")    public String getteacherList(@RequestBody String time){
           id++;        System.out.println(time);    //    System.out.println(nowtime.getId());     Map
   
     jsonMap = JSON.parseObject(time);        System.out.println(jsonMap.get("total"));        LocalDate date = LocalDate.now();        System.out.println(date);        Nowtime nowtime = new Nowtime();        nowtime.setNowtime(date.toString());        String ns=jsonMap.get("timenot").toString();        String totaltime=jsonMap.get("total").toString();        String begintime = jsonMap.get("begintime").toString();        nowtime.setTimenot(ns);        nowtime.setTotal(totaltime);        nowtime.setId(id);        nowtime.setBegintime(begintime);        timeDao.addtime(nowtime);        return "ok";        //return timenot;    }    @GetMapping("/gettime")    public String getalltime(){
           System.out.println("time!");     List
    
      nowtimes = timeDao.getall();        HashMap
     
       res = new HashMap<>();        res.put("data",nowtimes);        String users_json = JSON.toJSONString(res);        return users_json;    }}
     
    
   

xml语句：

   
       
    
             insert into data1.gettime(timenot,total,nowtime,begintime) values (#{timenot},#{total},#{nowtime},#{begintime});    
        
            SELECT * FROM gettime                    WHERE nowtime like #{nowtime}            
   

python的主程序：

import base64from predict import class_namesimport torchfrom torchvision import datasets, models, transformsimport cv2import numpy as npimport requestsfrom flask import Flask,make_response, jsonifyimport flaskfrom flask_cors import CORSimport socketimport threadingimport jsonimport osfrom io import BytesIOfrom multiprocessing import Processimport iofrom PIL import Image# 配置全局appapp = Flask(__name__)# 导入index中定义的所有函数#from autotrade.server.index import *def run_index():    # 启动web服务器，使用多线程方式，接收所有http请求    app.run(host='0.0.0.0', port=5000, threaded=True)def make_new_response(data):    res = make_response(jsonify({
   'code': 0, 'data': data}))    res.headers['Access-Control-Allow-Origin'] = '*'    res.headers['Access-Control-Allow-Method'] = '*'    res.headers['Access-Control-Allow-Headers'] = '*'    return resdef decode_base64(data):    """Decode base64, padding being optional.    :param data: Base64 data as an ASCII byte string    :returns: The decoded byte string.    """    missing_padding = len(data) % 4    if missing_padding != 0:        data += b'='* (4 - missing_padding)    # return base64.decodestring(data)    return base64.b64decode(data)@app.route("/test")def test():    res = "{'no':'dddd'}"    return make_new_response(res)CORS(app, resources=r'/*', supports_credentials=True)basedir = os.path.abspath(os.path.dirname(__file__))transform=transforms.Compose([            transforms.Resize(224),            transforms.CenterCrop(224),            transforms.ToTensor(),            transforms.Normalize(mean=[0.485,0.456,0.406],                                 std=[0.229,0.224,0.225])                            ])@app.route('/getpic', methods=['POST'])def getpic():    data = json.loads(flask.request.get_data("data"))    data_64 = str.encode(data['data'])    print(type(data_64))    #print(data_64)    print('------------------------')    print(str(data_64, 'utf8'))    imgdata = decode_base64(data_64)    file = open('1.jpg', 'wb')    file.write(imgdata)    file.close()    image = Image.open(r"1.jpg").convert('RGB')    image = transform(image).unsqueeze(0)    modelme = torch.load('modefresnet.pkl')    modelme.eval()    outputs = modelme(image)    _, predict = torch.max(outputs.data, 1)    for j in range(image.size()[0]):        print('predicted: {}'.format(class_names[predict[j]]))    return class_names[predict[j]] if __name__ == "__main__":    app.run(debug=True)

深度学习的处理图片的网络模型就不贴了，需要的可以留言

用的是Resnet残差网络。识别速度还是很快的，判断的正确率也比较高。（训练的数据集很少，只有六百多张）

##Y1BCojf69##4;1%yBNfY3ne6a!/

下例为从指定的层提取ResNet50的特征。

import torch

from torch import nn import torchvision.models as models import torchvision.transforms as transforms import cv2

class FeatureExtractor(nn.Module): # 提取特征工具

def init(self, submodule, extracted_layers): super(FeatureExtractor, self).init() self.submodule = submodule self.extracted_layers = extracted_layers

def forward(self, x):    outputs = []    for name, module in self.submodule._modules.items():        if name is "fc":             x = x.view(x.size(0), -1)        x = module(x)        if name in self.extracted_layers:            outputs.append(x)    return outputs

model = models.resnet50(pretrained=True) # 加载resnet50工具

model = model.cuda() model.eval()

img=cv2.imread(‘test.jpg’) # 加载图片

img=cv2.resize(img,(224,224)); img=cv2.cvtColor(img,cv2.COLOR_BGR2RGB) transform = transforms.Compose( [transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))]) img=transform(img).cuda() img=img.unsqueeze(0)

model2 = FeatureExtractor(model, [‘layer3’]) # 指定提取 layer3 层特征

with torch.no_grad(): out=model2(img) print(len(out), out[0].shape)

转载地址：http://lvten.baihongyu.com/