Noise Segmentation In Unet

I am using Monai in pytorch for 2D segmentation. The output of the algorithm is showing a dice score accuracy of over 80 percent yet the images look like plain noise.

The parameters for the Unet:
monai.networks.nets.UNet(2, 1, 2, channels= (32, 64, 128, 256, 512), strides=(2, 2, 2, 2))

I am not sure what is causing this issue.