基于口罩佩戴公开赛数据实践口罩佩戴识别

这是之前网上举办的口罩佩戴公开赛的数据集，我周末正好有时间就想着拿来做一个口罩佩戴识别系统，跟我前面的一篇博文相互照应。链接在下面，感兴趣可以看下：

                              《助力防疫，基于安防摄像头的人脸佩戴口罩检测》 

这里做的不是目标检测，而是图像识别任务。

效果如下：

 网上公开赛的数据集下载解压缩后如下：

 sample_submit.csv是提交样例格式，这里可以直接删掉。

进入train目录，如下：

 可见：大赛给定的数据集一共是划分了三个类别，从上到下分别表示：未正确佩戴口罩、佩戴口罩、未佩戴口罩。

我自己开发不打算分成三个类别，这里将未正确佩戴口罩划归到佩戴口罩类别下，当然了这个本身无所谓的，我只是个人做法而已。

首先解析数据集：

def parseData2H5():
    '''
    加载本地数据集创建H5数据集
    '''
    X_train, y_train =[], []
    X_test, y_test =[], []
    picDir='data/train/'
    for one_label in os.listdir(picDir):
        for one_pic in os.listdir(picDir+one_label+'/'):
            if one_pic.endswith("jpg") or one_pic.endswith("png") or one_pic.endswith("jpeg"):
                try:
                    one_path = picDir+one_label+'/' + one_pic
                    print("one_path: ", one_path)
                    #图片
                    one_img = cv2.imread(one_path)
                    one_img = cv2.resize(one_img,(100,100))
                    one_img = one_img.transpose((2,0,1))
                    #标签
                    one_pic_classes = one_label
                    one_y = getY(one_pic_classes)
                    #整合
                    X_train.append(one_img)
                    y_train.append(one_y)
                except Exception as e:
                    print("train Exception: ", e)
    X_train = np.array(X_train)
    picDir='data/test/'
    for one_label in os.listdir(picDir):
        for one_pic in os.listdir(picDir+one_label+'/'):
            if one_pic.endswith("jpg") or one_pic.endswith("png") or one_pic.endswith("jpeg"):
                try:
                    one_path = picDir+one_label+'/' + one_pic
                    print("one_path: ", one_path)
                    #图片
                    one_img = cv2.imread(one_path)
                    one_img = cv2.resize(one_img,(100,100))
                    one_img = one_img.transpose((2,0,1))
                    #标签
                    one_pic_classes = one_label
                    one_y = getY(one_pic_classes)
                    #整合
                    X_test.append(one_img)
                    y_test.append(one_y)
                except Exception as e:
                    print("test Exception: ", e)
    X_test = np.array(X_test)
    f = h5py.File("dataset.h5")
    f['X_train'] = X_train
    f['X_test'] = X_test
    f['y_train'] = y_train
    f['y_test'] = y_test
    print("y_train_0: ", y_train[0])
    print("y_test_0: ", y_test[0])
    f.close()

之后构建模型：

def initModel(h=224, w=224, way=3):
    """
    构建模型
    """
    model = Sequential()
    input_shape = (h, w, way)
    model.add(Conv2D(64, (3, 3), input_shape=input_shape))
    model.add(Activation("relu"))
    model.add(Dropout(0.3))
    model.add(Conv2D(64, (3, 3)))
    model.add(Activation("relu"))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Flatten())
    model.add(Dense(1024))
    model.add(Activation("relu"))
    model.add(Dropout(0.3))
    model.add(Dense(numbers))
    model.add(Activation("sigmoid"))
    lrate = 0.01
    decay = lrate / 100
    sgd = SGD(lr=lrate, momentum=0.9, decay=decay, nesterov=False)
    model.compile(loss="binary_crossentropy", optimizer=sgd, metrics=["accuracy"])
    print(model.summary())
    return model

训练模型就是正常的流程了，这里就不在介绍了，在我之前的文章里面有很多图像识别的项目，这里就不再赘述细节了。训练过程可视化如下：

 这里开发了批量测试函数用于测试整体的识别精度：

def batchTest(dataDir="data/test/"):
    '''
    测试集测试
    '''
    count_dict={}
    for one_label in os.listdir(dataDir):
        print("one_label: ", one_label)
        count_dict[one_label]={}
        count_dict[one_label]["right"]=0
        count_dict[one_label]["wrong"]=0
        oneDir=dataDir+one_label+"/"
        for one_pic in os.listdir(oneDir):
            one_path=oneDir+one_pic
            one_pred,one_proba=singleTest(image_path=one_path)
            if one_pred==one_label:
                count_dict[one_label]["right"]+=1
            else:
                count_dict[one_label]["wrong"]+=1
    print("count_dict: ", count_dict)
    with open("count_dict.json","w") as f:
        f.write(json.dumps(count_dict))
    acc_dict={}
    for one_label in count_dict:
        one_right,one_wrong=count_dict[one_label]["right"],count_dict[one_label]["wrong"]
        one_acc=one_right/(one_right+one_wrong)
        print("one_label: ", one_label)
        print("one_acc: ", one_acc)
        acc_dict[one_label]=one_acc
    print("acc_dict: ", acc_dict)

感兴趣可以直接使用，也可以基于自己的个性化应用进行改造。