YOLO-V3实时检测实现（opencv+python实现）——改进——＞更加的易懂

• • 1.上一篇文章YOLO-V3实时检测实现
  • 2.上对于上一篇的改进之处
  • 3.代码实战
  • • （1）网络文件(.cfg)和权重文件的读取
    • （2）类别文件的读取
    • （3）网络模型的加载
    • （4）得到预测的结果
    • （5）绘制图形
    • （6）整体的代码
    • （7）单张图片的预测
    • （8）实时的检测

https://blog.csdn.net/Keep_Trying_Go/article/details/125503943

#获取网络模型
    model=cv2.dnn_DetectionModel(net)
    #设置网络的输入参数
    model.setInputParams(scale=1/255,size=(416,416))
    #进行预测
    class_id,scores,boxes=model.detect(frame,confThreshold=Confidence_thresh,
                 nmsThreshold=Nms_thresh)
    #返回预测的类别和坐标
    return class_id,scores,boxes

不需要存储预测框坐标的转换，需要进行相关 （坐标，高宽，置信度，概率值以及类别）进行重新的遍历和存储，直接得到输出的结果。

#读取YOLO-V3权重文件和网络配置文件
net=cv2.dnn.readNet(model='dnn_model/yolov3.weights',config='dnn_model/yolov3.cfg')

#设置置信度阈值和非极大值抑制的阈值
Confidence_thresh=0.2
Nms_thresh=0.35
#读取coco.names文件中的类别
with open('dnn_model/coco.names','r') as fp:
    classes=fp.read().splitlines()

#yolo-v3检测
def detect(frame):
    #获取网络模型
    model=cv2.dnn_DetectionModel(net)
    #设置网络的输入参数
    model.setInputParams(scale=1/255,size=(416,416))
    #进行预测
    class_id,scores,boxes=model.detect(frame,confThreshold=Confidence_thresh,
                 nmsThreshold=Nms_thresh)
    #返回预测的类别和坐标
    return class_id,scores,boxes

#进行预测
        class_ids,scores,boxes=detect(frame)

#绘制矩形框
        for (class_id,box) in enumerate(boxes):
            (x,y,w,h)=box
            class_name=classes[class_ids[class_id]]
            confidence=scores[class_id]
            confidence=str(round(confidence,2))
            cv2.rectangle(img=frame,pt1=(x,y),pt2=(x+w,y+h),
                          color=(0,255,0),thickness=2)
            text=class_name+' '+confidence
            cv2.putText(img=frame,text=text,
                        org=(x,y-10),fontFace=cv2.FONT_HERSHEY_SIMPLEX,
                        fontScale=1.0,color=(0,255,0),thickness=2)

import os
import cv2
import numpy as np

#创建窗口
# cv2.namedWindow(winname='detect',flags=cv2.WINDOW_AUTOSIZE)
# cv2.resizeWindow(winname='detect',width=750,height=600)

#读取YOLO-V3权重文件和网络配置文件
net=cv2.dnn.readNet(model='dnn_model/yolov3.weights',config='dnn_model/yolov3.cfg')

#设置置信度阈值和非极大值抑制的阈值
Confidence_thresh=0.2
Nms_thresh=0.35
#读取coco.names文件中的类别
with open('dnn_model/coco.names','r') as fp:
    classes=fp.read().splitlines()

#yolo-v3检测
def detect(frame):
    #获取网络模型
    model=cv2.dnn_DetectionModel(net)
    #设置网络的输入参数
    model.setInputParams(scale=1/255,size=(416,416))
    #进行预测
    class_id,scores,boxes=model.detect(frame,confThreshold=Confidence_thresh,
                 nmsThreshold=Nms_thresh)
    #返回预测的类别和坐标
    return class_id,scores,boxes


#实时的检测
def detect_time():
    #开启摄像头 'video/los_angeles.mp4' or 'video/soccer.mp4'
    cap=cv2.VideoCapture(0)

    while cap.isOpened():
        OK,frame=cap.read()
        if not OK:
            break

        frame=cv2.flip(src=frame,flipCode=2)
        # frame=cv2.resize(src=frame,dsize=(416,416))
        #进行预测
        class_ids,scores,boxes=detect(frame)
        #绘制矩形框
        for (class_id,box) in enumerate(boxes):
            (x,y,w,h)=box
            class_name = classes[class_ids[class_id]]
            confidence = scores[class_id]
            confidence=str(round(confidence,2))
            cv2.rectangle(img=frame,pt1=(x,y),pt2=(x+w,y+h),
                          color=(0,255,0),thickness=2)
            text=class_name+' '+confidence
            cv2.putText(img=frame,text=text,
                        org=(x,y-10),fontFace=cv2.FONT_HERSHEY_SIMPLEX,
                        fontScale=1.0,color=(0,255,0),thickness=2)

        cv2.imshow('detect',frame)
        key=cv2.waitKey(1)
        if key==27:
            break

    cap.release()
#单张图片的检测
def signal_detect(image_path='images/face1.jpg'):
    frame=cv2.imread(image_path)
    frame = cv2.resize(src=frame, dsize=(416, 416))
    # 进行预测
    class_ids, scores, boxes = detect(frame)
    # 绘制矩形框
    for (class_id, box) in enumerate(boxes):
        (x, y, w, h) = box
        class_name = classes[class_ids[class_id]]
        confidence = scores[class_ids[class_id]]
        confidence = str(round(confidence, 2))
        cv2.rectangle(img=frame, pt1=(x, y), pt2=(x + w, y + h),
                      color=(0, 255, 0), thickness=2)
        text = class_name + ' ' + confidence
        cv2.putText(img=frame, text=text,
                    org=(x, y - 10), fontFace=cv2.FONT_HERSHEY_SIMPLEX,
                    fontScale=1.0, color=(0, 255, 0), thickness=2)

    cv2.imshow('detect', frame)
    cv2.waitKey(0)

cv2.destroyAllWindows()


if __name__ == '__main__':
    print('Pycharm')
    # signal_detect()
    detect_time()

yolov3视频演示1