Unet- Convolutional Networks for Biomedical Image Segmentation

Idea

This is a paper implementation of the UNET model in Pytorch

As you can see from the above figure, the model looks like a U shape (hence the name). Most important thing to notice here is that the Network is a simple Encoder decoder Network but with skip connections as well. Also the block of 3 CONV->ReLU always gets repeated which is accompanied by a Max-Pooling layer.

This is more of a vanilla implementation with no fancy stuff. I just implemented this in here to try out with pytorch.

If you want more information on the orginal paper, you can definitely check it out here.

Ways in which this can be used

• You can use this as a starting stage while working on a Unet project
• You can refer to this if you want to work with implementing an paper (Always start off with the architecture. It's the easiest).