Error Message:
TypeError: 未找到必需的参数'rec' (位置 2)
Source Code:
ori_img = img[0].detach().cpu().numpy()
ori_img = (ori_img*127.5 + 127.5).transpose([1,2,0]).astype(np.uint8)
x1, y1, x2, y2 = int(bb[0]), int(bb[1]), int(bb[2]), int(bb[3])
ori_img = cv2.rectangle(img=ori_img, pt1=(x1, y1), pt2=(x2, y2), color=(255, 255, 0), thickness=2)
The above error occurred, and after searching online, 99% of the solutions suggest converting all vertex coordinates to int type. However, in my case, I have already done the conversion, but the problem still persists.
At the same time, ori_img has also been converted to uint8 type, which should be fine. But the problem actually lies with ori_img.
The following code is the modified version, with an additional line that calls the function np.ascontiguousarray. After searching online, it is explained as:
Converts an array with non-contiguous memory storage to an array with contiguous memory storage, making it faster to run.
It is not difficult to understand the literal meaning, but I don't quite understand why a numpy matrix that has been converted from a Tensor and all the way down would become non-contiguous in memory.
ori_img = img[0].detach().cpu().numpy()
ori_img = (ori_img*127.5 + 127.5).transpose([1,2,0]).astype(np.uint8)
ori_img = np.ascontiguousarray(ori_img)
x1, y1, x2, y2 = int(bb[0]), int(bb[1]), int(bb[2]), int(bb[3])
ori_img = cv2.rectangle(img=ori_img, pt1=(x1, y1), pt2=(x2, y2), color=(255, 255, 0), thickness=2)