U-net is a very popular image segmentation architecture that produces segmentation masks. Could anyone please clarify to me what is the difference between U-net and another V-net architecture? What can I do with these architectures for segmenting my volumetric data?
Some interesting papers dealing with that would be:
1- Deep Learning Techniques for Medical Image Segmentation: Achievements and Challenges
Mohammad Hesam Hesamian, Wenjing Jia,1 Xiangjian He, and Paul Kennedy
2- A survey on U-shaped networks in medical image segmentations
U refers to the shape of the network and V refers to size of the input (volumetric input). Therefore, the architecture is very close, but the difference is in input size. For your problem, you need volumetric input. Also, you have to tune the architecture for your dataset.
In the following link you can see the "U" shaped structure of U-NET formed by encoder and decoder stage: https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/. In addition, the original U-net approach was applied and tested on highly challenging 2D datasets (Phase contrast and DIC microscopy data in the 2015 ISBI challenge). So some changes in terms of the U-net architecture might be required for segmenting in 3D data.
Some interesting papers dealing with that would be:
1- Deep Learning Techniques for Medical Image Segmentation: Achievements and Challenges
Mohammad Hesam Hesamian, Wenjing Jia,1 Xiangjian He, and Paul Kennedy
2- A survey on U-shaped networks in medical image segmentations
For medical image processing U-net architecture is best suited and easy to implement using python and the results are more accurate as compared to other algorithms
Dear reserachers
- Deep Learning Techniques for Medical Image Segmentation: Achievements and Challenges
Regards
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation
- 3D