解释 readNetFromCaffe 的 forward 方法返回的值答案

【问题标题】：Explain the values returned by readNetFromCaffe's forward method解释 readNetFromCaffe 的 forward 方法返回的值
【发布时间】：2020-12-16 01:52:07
【问题描述】：

net = cv2.dnn.readNetFromCaffe(prototxt, model)
detections = net.forward()

检测将具有形状为 (1, 1, 200, 7) 的 4D 数组。那里有什么不同的价值观？

    for i in range(0, detections.shape[2]):
        confidence = detections[0, 0, i, 2]

上面的循环在第 3 维获取置信度值，提供的是行号，第 4 维是列号，因此很明显，物体以如此高的置信度或概率被检测到。但我无法理解其他参数。

        box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])

要创建一个框，上面的代码用于 3 到 6 列及其值。那么第 2 到第 6 列中的这些值是什么？

下面的代码重现值...

from imutils.video import VideoStream
import imutils
import numpy as np
import cv2
import argparse
import time

# construct the argument parser
ap = argparse.ArgumentParser()
ap.add_argument("-p", "--prototxt", required=True, help="path to caffe deploy prototxt file")
ap.add_argument("-m", "--model", required=True, help="path to pre-trained caffe model")
ap.add_argument("-c", "--confidence", type=float, default=0.5, help="minimum confidence required")
args = vars(ap.parse_args())

# MODEL - load the model which will be used to predict
print("[INFO] loading model...")
net = cv2.dnn.readNetFromCaffe(args["prototxt"], args["model"])

# INPUT - start video to capture frames
print("[INFO] starting video stream...")
vs = VideoStream(src=0).start()
# vs = VideoStream(usePiCamera=True).start()  # this is to stream video for Raspberry Pi camera
# vs = FileVideoStream(path='/path to file')  # this is to get video content from file
time.sleep(2)  # allow the cam to warm up

# for live streaming, while loop will be required to capture frames
while True:
    frame = vs.read()
    frame = imutils.resize(frame, width=400)

    # get the dimensions of the frame
    print('shape of frame:', frame.shape)
    (h, w) = frame.shape[:2]

    # blob the frame
    blob = cv2.dnn.blobFromImage(frame, scalefactor=1.0,
                                 size=(300, 300), mean=(104.0, 177.0, 123.0))

    # use the blob for detection
    net.setInput(blob)
    detections = net.forward()
    print(detections)
    # loop over the detections
    for i in range(0, detections.shape[2]):
        confidence = detections[0, 0, i, 2]
        if confidence < args['confidence']:
            continue

        # create box
        box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
        (startX, startY, endX, endY) = box.astype("int")

        # draw the box on the face
        text = "{:.2f}%".format(confidence * 100)
        y = startY - 10 if startY - 10 > 10 else startY + 10
        cv2.rectangle(frame, (startX, startY), (endX, endY), (0, 255, 0), thickness=2)
        cv2.putText(frame, text, (startX, y), cv2.FONT_HERSHEY_SIMPLEX, 0.45, (0, 0, 255), 2)

    # show on screen
    cv2.imshow("frame", frame)
    if cv2.waitKey(1) & 0xFF == 27:
        break

cv2.destroyAllWindows()
vs.stop()

【问题讨论】：

标签： python opencv caffe face-detection

【解决方案1】：

第一列代表它所属的类id 第二列代表置信度值，通过将其与置信度阈值进行比较，帧中是否有任何对象 [3:7] 即 3,4,5,6 列代表在对象周围创建框的坐标/点，它们存储为 numpy 数组，因此可以转换为 int 类型以制作框。

【讨论】：