使用 TF 和 OpenCV 读取参数“img”的预期 Ptr<cv::UMat>

Question

我有来自here 的这段代码，对here 做了一些修改。

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


# construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()

ap.add_argument("-c", "--confidence", type=float, default=0.8,
    help="minimum probability to filter weak detections")
args = vars(ap.parse_args())


classes_90 = [ "person", "bicycle", "car", "motorcycle",
            "airplane", "bus", "train", "truck", "boat", "traffic light", "fire hydrant",
            "unknown", "stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse",
            "sheep", "cow", "elephant", "bear", "zebra", "giraffe", "unknown", "backpack",
            "umbrella", "unknown", "unknown", "handbag", "tie", "suitcase", "frisbee", "skis",
            "snowboard", "sports ball", "kite", "baseball bat", "baseball glove", "skateboard",
            "surfboard", "tennis racket", "bottle", "unknown", "wine glass", "cup", "fork", "knife",
            "spoon", "bowl", "banana", "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog",
            "pizza", "donut", "cake", "chair", "couch", "potted plant", "bed", "unknown", "dining table",
            "unknown", "unknown", "toilet", "unknown", "tv", "laptop", "mouse", "remote", "keyboard",
            "cell phone", "microwave", "oven", "toaster", "sink", "refrigerator", "unknown",
            "book", "clock", "vase", "scissors", "teddy bear", "hair drier", "toothbrush" ] 
# Leemos las clases disponibles en openImages
CLASSES = classes_90  #New list of classess with 90 classess.
print(CLASSES)

# Le damos colores a las cajas para cada clase
COLORS = np.random.uniform(0, 255, size=(len(CLASSES), 3)) 

# Importamos el modelo de red
cvNet = cv2.dnn.readNetFromTensorflow('faster_rcnn_inception_v2_coco_2018_01_28/frozen_inference_graph.pb', 'faster_rcnn_inception_v2_coco_2018_01_28/resnet.pbtxt')

# Leemos una imagen
img = cv2.VideoCapture('people.mp4')  


while img.isOpened():
    ret, frame = img.read()
    
    if not ret:
        
   

 break


#img = cv2.imread(args["image"])

# Obtenemos las dimensiones de la imagen
h = frame.shape[0] # Alto
w = frame.shape[1] # Ancho
img = np.array(img)
cvNet.setInput(cv2.dnn.blobFromImage(img, size=(h, w), swapRB=True, crop=False))
detections = cvNet.forward()


# loop over the detections
for i in np.arange(0, detections.shape[2]):
    # extract the confidence (i.e., probability) associated with
    # the prediction
    confidence = detections[0, 0, i, 2]

    # filter out weak detections by ensuring the `confidence` is
    # greater than the minimum confidence
    if confidence > args["confidence"]:
        # extract the index of the class label from the
        # `detections`, then compute the (x, y)-coordinates of
        # the bounding box for the object
        idx = int(detections[0, 0, i, 1])
        print(idx   )
        box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
        (startX, startY, endX, endY) = box.astype("int")

        # draw the prediction on the frame
        label = "{}: {:.2f}%".format(CLASSES[idx],
            confidence * 100)
        cv2.rectangle(img, (startX, startY), (endX, endY),
            COLORS[idx], 2)
        y = startY - 15 if startY - 15 > 15 else startY + 15
        cv2.putText(img, label, (startX, y),
            cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLORS[idx], 2)

        print(label)


out_img = cv2.resize(img, (640, 480))
out.write(out_img)
cv2.imshow('img', img)
#cv2.waitKey()
if cv2.waitKey(25) & 0xFF == ord('q'):

    cv2.destroyAllWindows()
    cap.release()
    out.release()

并收到此错误 Expected Ptr<cv::UMat> for argument 'img'，在查看了该问题的大多数可用解决方案后，似乎起初输入不是数组，因此更改为 np.array 但不起作用，打印图像显示图像这是视频中的一帧，所以图像就在那里。

因此，我无法弄清楚究竟是什么导致了这个问题。如果您使用 cv2.imread() 仅传递一个图像，添加此代码也可以正常工作。

Answer 1

我错误地将帧传递给数组，我的意思是我正在创建一个包含一些不同变量的空数组，而不是从 cv2.VideoCapture() 方法获得的帧。

以下是更新后的代码：

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


# construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()

ap.add_argument("-c", "--confidence", type=float, default=0.8,
    help="minimum probability to filter weak detections")
args = vars(ap.parse_args())


classes_90 = [ "person", "bicycle", "car", "motorcycle",
            "airplane", "bus", "train", "truck", "boat", "traffic light", "fire hydrant",
            "unknown", "stop sign", "parking meter", "bench", "bird", "cat", "dog", "horse",
            "sheep", "cow", "elephant", "bear", "zebra", "giraffe", "unknown", "backpack",
            "umbrella", "unknown", "unknown", "handbag", "tie", "suitcase", "frisbee", "skis",
            "snowboard", "sports ball", "kite", "baseball bat", "baseball glove", "skateboard",
            "surfboard", "tennis racket", "bottle", "unknown", "wine glass", "cup", "fork", "knife",
            "spoon", "bowl", "banana", "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog",
            "pizza", "donut", "cake", "chair", "couch", "potted plant", "bed", "unknown", "dining table",
            "unknown", "unknown", "toilet", "unknown", "tv", "laptop", "mouse", "remote", "keyboard",
            "cell phone", "microwave", "oven", "toaster", "sink", "refrigerator", "unknown",
            "book", "clock", "vase", "scissors", "teddy bear", "hair drier", "toothbrush" ] 
# Leemos las clases disponibles en openImages
CLASSES = classes_90  #New list of classess with 90 classess.
print(CLASSES)

# Le damos colores a las cajas para cada clase
COLORS = np.random.uniform(0, 255, size=(len(CLASSES), 3)) 

# Importamos el modelo de red
cvNet = cv2.dnn.readNetFromTensorflow('faster_rcnn_inception_v2_coco_2018_01_28/frozen_inference_graph.pb', 'faster_rcnn_inception_v2_coco_2018_01_28/resnet.pbtxt')

# Leemos una imagen
img = cv2.VideoCapture('people.mp4')  


while img.isOpened():
    ret, frame = img.read()
    
    if not ret:
        
   

 break


#img = cv2.imread(args["image"])

# Obtenemos las dimensiones de la imagen
h = frame.shape[0] # Alto
w = frame.shape[1] # Ancho
img = np.array(frame)
cvNet.setInput(cv2.dnn.blobFromImage(img, size=(h, w), swapRB=True, crop=False))
detections = cvNet.forward()


# loop over the detections
for i in np.arange(0, detections.shape[2]):
    # extract the confidence (i.e., probability) associated with
    # the prediction
    confidence = detections[0, 0, i, 2]

    # filter out weak detections by ensuring the `confidence` is
    # greater than the minimum confidence
    if confidence > args["confidence"]:
        # extract the index of the class label from the
        # `detections`, then compute the (x, y)-coordinates of
        # the bounding box for the object
        idx = int(detections[0, 0, i, 1])
        print(idx   )
        box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
        (startX, startY, endX, endY) = box.astype("int")

        # draw the prediction on the frame
        label = "{}: {:.2f}%".format(CLASSES[idx],
            confidence * 100)
        cv2.rectangle(img, (startX, startY), (endX, endY),
            COLORS[idx], 2)
        y = startY - 15 if startY - 15 > 15 else startY + 15
        cv2.putText(img, label, (startX, y),
            cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLORS[idx], 2)

        print(label)


out_img = cv2.resize(img, (640, 480))
out.write(out_img)
cv2.imshow('img', img)
#cv2.waitKey()
if cv2.waitKey(25) & 0xFF == ord('q'):

    cv2.destroyAllWindows()
    cap.release()
    out.release()