下述两个函数均可：
1.
def image_Tensor2ndarray(image_tensor: torch.Tensor):
"""
将tensor转化为cv2格式
"""
assert (len(image_tensor.shape) == 4 and image_tensor.shape[0] == 1)
# 复制一份
image_tensor = image_tensor.clone().detach()
# 到cpu
image_tensor = image_tensor.to(torch.device('cpu'))
# 反归一化
# input_tensor = unnormalize(input_tensor)
# 去掉批次维度
image_tensor = image_tensor.squeeze()
# 从[0,1]转化为[0,255]，再从CHW转为HWC，最后转为cv2
image_cv2 = image_tensor.mul_(255).add_(0.5).clamp_(0, 255).permute(1, 2, 0).type(torch.uint8).numpy()
# RGB转BRG
# image_cv2 = cv2.cvtColor(image_cv2, cv2.COLOR_BGR2RGB)
return image_cv2
def image_Tensor2np(image_tensor: torch.Tensor):
"""
将tensor转化为cv2格式
"""
assert (len(image_tensor.shape) == 4 and image_tensor.shape[0] == 1)
# 复制一份
image_tensor = image_tensor.cpu().detach()
# 去掉批次维度
image_tensor = image_tensor.squeeze()
# 将tensor数据转为numpy数据
image_tensor = image_tensor.numpy()
# 反归一化normalize，将图像数据扩展到[0,255]
maxValue = image_tensor.max()
image_tensor = image_tensor * 255 / maxValue
# float32-->uint8
image_cv2 = np.uint8(image_tensor)
# 再从CHW转为HWC
image_cv2 = image_cv2.transpose(1, 2, 0)
# # RGB转BRG
# image_cv2 = cv2.cvtColor(image_cv2, cv2.COLOR_RGB2BGR)
return image_cv2
