数人 yolo 技术笔记

数人 2

opencv自带的Haar特征分类器进行人脸检测 效果太差，尝试使用 yolo看看
从GitHub上下载了yolov3.cfg、yolov3.weights、coco.names，然后试着数人

import cv2  `
import numpy as np  
  
# 加载YOLO网络配置和权重  
net = cv2.dnn.readNet("d:\\yolov3.weights", "d:\\yolov3.cfg")  
classes = []  
with open("d:\\coco.names", "r") as f:  
    classes = [line.strip() for line in f.readlines()]  
  
# 加载图像  
img = cv2.imread("d:\\202408212.jpg")  
img = cv2.resize(img, None, fx=0.4, fy=0.4)  
height, width, channels = img.shape  
  
# 检测图像中的对象  
blob = cv2.dnn.blobFromImage(img, 0.00392, (416, 416), (0, 0, 0), True, crop=False)  
net.setInput(blob)  
outs = net.forward(net.getUnconnectedOutLayersNames())  
  
# 显示信息在屏幕上  
class_ids = []  
confidences = []  
boxes = []  
for out in outs:  
    for detection in out:  
        scores = detection[5:]  
        class_id = np.argmax(scores)  
        confidence = scores[class_id]  
        if confidence > 0.5:  
            # 对象检测  
            center_x = int(detection[0] * width)  
            center_y = int(detection[1] * height)  
            w = int(detection[2] * width)  
            h = int(detection[3] * height)  
  
            # 矩形框  
            x = int(center_x - w / 2)  
            y = int(center_y - h / 2)  
  
            boxes.append([x, y, w, h])  
            confidences.append(float(confidence))  
            class_ids.append(class_id)  
  
# 使用NMS去除多余的框  
indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)  
  
# 计算人数  
person_count = 0  
color = (255, 0, 0)  
for i in range(len(boxes)):  
    if i in indexes:  
        x, y, w, h = boxes[i]  
        label = str(classes[class_ids[i]])  
        if label == 'person':  # 假设'person'是类名中表示人的标签  
            person_count += 1  
            cv2.rectangle(img, (x, y), (x + w, y + h), color, 1)  
            #cv2.putText(img, label, (x, y - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 1) 
  
print(f"Detected {person_count} people in the image.")

  
# 显示图像  
plt.figure(figsize=(20, 16))  
plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))  # 将BGR转换为RGB以正确显示颜色  
plt.show()

效果好一些了，正面、没有被遮掩的都可以计算到了