Skip to content

MEDICAL IMAGE SEGMENTATION USING DEFORMABLE ATTENTION (DEEP LEARNING) FRAMEWORK

Talk Abstract

Medical Image Segmentation is a challenging area of research where Deep Learning algorithms have shown lot of success in recent times. In this talk, we would get an introduction to Deep Learning based segmentation of Medical images. A brief overview of UNets would be covered to introduce the audience to segmentation architectures.

The second part of the talk would be on Attention mechanism. Attention mechanism has revolutionized Deep Learning algorithms, especially in NLP domain and more recently in Computer Vision tasks.

In this talk, after a brief introduction to attention blocks (from NLP context) we start with some of the attention approaches for computer vision tasks. We dive in more detail on couple of attention blocks – criss-cross attention and attention-augmented-convolution. Some of the active frontiers in attention blocks (deformable attention) would be finally discussed to give some insights on current research directions.

Presented By

Kumar Rajamani

Senior Manager Algorithm | KLA Tencor