UCTransNet

The U-Net model (Ronneberger et al) is a used for semantic segmentation applied to the medical imaging domain.

UCTransnet : Rethinking the Skip Connections in U-Net from a Channel-wise Perspective with Transformer () implements a mechanism to balance the contributions of the skip connections of the original model.

This paper has been presented on May 18th 2020 at the CDLF.

Paper

The publication can be found on Arxiv.

Slides

• Pdf version of the slides - UCTRANSNET: RETHINKING THE SKIP CONNECTIONS IN U-NET FROM A CHANNEL-WISE PERSPECTIVE WITH TRANSFORMER, PRESENTED BY TED EDMONDS

Key concepts

• Encoder / Decoder
• Attention mechanism
• Embedding
• Instance normalization

Code implementation

The official implementation of this model is available on Github.

Reference

Paper with code