As shown below: lateral_map_2 is a tensor type vector (prediction map) generated by the network, and now I want to convert it into a picture, and the personal test is effective.
from torchvision.utils import save_image
res = lateral_map_2 # [1,1,352,352]
# res = res * 255
res.cuda()
tensor_float = res.float()
img = tensor_float[0]
# (图片,存放图片的路径)
save_image(img, opt.save_path + name)