正在尝试使用ssd_inception_v2_coco_2018_01_28在图像上实现对象检测模型。我试图获取顶部带有边界框的输出。
1)ssd_inception_v2_coco_2018_01_28对象检测模型。已成功转换为IR。
2)拍摄输入图像并根据模型的输入层信息进行预处理。
3)成功提取模型输出。
4)将边界框添加到输出图像。
但是出现以下错误,请帮忙!
源代码
#Preprocessing image
img = r'D:\SriSailam.jpg'
img = cv2.imread(img)
img = cv2.resize(img,(300,300))
img = img.transpose((2,0,1))
img = img.reshape(1,3,300,300)
exec_net = engine.load_network(net,'cpu')
#Asynchronous request
infer_request_handle = exec_net.start_async(request_id=0, inputs={'image_tensor': img})
infer_status = infer_request_handle.wait()
output = next(iter(net.outputs))
n,c,h,w = img.shape
result = infer_request_handle.outputs[output] #output layer shape
'''
Draw bounding boxes onto the image.
'''
for box in result[0][0]: # Output shape is 1x1x100x7
conf = box[2] #box[2] = result[0][0][2] is a probability value
print("probability value : ",conf)
print("box[0] : ",box[0])
print("box[1] : ",box[1])
print("xmin : ", box[3])
print("ymin : ", box[4])
print("xmax : ", box[5])
print("ymax : ", box[6])
if conf >= 0:
xmin = int(box[3] * w) #box[3] - xmin
ymin = int(box[4] * h) #box[4] - ymin
xmax = int(box[5] * w) #box[5] - xmax
ymax = int(box[6] * h) #box[6] - ymax
rect = cv2.rectangle(img, ((xmin), (ymin)), ((xmax), (ymax)), (0, 0, 255), 1)
cv2.imshow(rect,'face_detection')
cv2.waitKey(0)
输出
C:\Users\adity\Desktop\openvino>py check_infer.py
probability value : 0.6344592
box[0] : 0.0
box[1] : 1.0
xmin : 0.1831626
ymin : 0.10697469
xmax : 0.86960804
ymax : 1.0
Traceback (most recent call last):
File "check_infer.py", line 68, in <module>
rect = cv2.rectangle(img, ((xmin), (ymin)), ((xmax), (ymax)), (0, 0, 255), 1)
TypeError: an integer is required (got type tuple)
复制了您的错误。在此版本的OpenCV(4.2.0)中,错误消息非常令人困惑。
如果您在OpenCV 4.3.0中执行相同的操作,则会得到更清晰的错误,例如
File "<stdin>", line 1, in <module>
TypeError: Expected Ptr<cv::UMat> for argument 'img'
基本上说您传递给img
的cv2.rectangle
参数的类型错误。问题是您致电
img = img.transpose((2,0,1))
img = img.reshape(1,3,300,300)
img
不能被OpenCV解释为图像。例如,可以通过以下方式解决此问题:1.转换前将img
复制到另一个变量imgForDisplaying
2.使用imgForDisplaying
绘制矩形。
img = r'D:\SriSailam.jpg'
img = cv2.imread(img)
img = cv2.resize(img,(300,300))
imgToDisplay = img
img = img.transpose((2,0,1))
img = img.reshape(1,3,300,300)
...
rect = cv2.rectangle(imgToDisplay, (xmin, ymin), (xmax, ymax), (0, 0, 255), 1)