Comments (6)
Thanks for pointing out this issue.
- About the missing code reference. We have added the code reference from (https://github.com/jeonsworld/ViT-pytorch).
- About the concurrent Arxiv. Our proposed TransUNet is inspired by ViT and U-Net (before the release date of https://arxiv.org/abs/2012.15840) in that ViT is powerful at extracting global contexts whereas U-Net is powerful at segmenting finer details. And we also want to note that one of our key points is the skip connection used in TransUNet (Sec. 4, Fig. 2), which makes our architecture quite different from this Arxiv reference. Lastly, we would like to highlight that we have already cited this Arxiv reference in our paper.
P.S. Please check http://cvpr2020.thecvf.com/submission/main-conference/author-guidelines for the definition of concurrent submission in case you are not aware of this. I wish you could educate yourself before posting comments next time.
from transunet.
@ching-sui1995 As a third-party who has done a similar architecture (transformer for segmentation), I'd like to participate in this discussion. My main points:
- First of all, you can find the repo of my own model (https://github.com/segtran/segtran). It's done in last July-Aug and had won a place in a competition. (This is not my main github account. As my paper is under review, I have to stay anonymous.)
- I believe quite a few teams were doing similar explorations concurrently last year, so I won't claim my model was the first, or earlier than TransUNet.
- All our works, no matter SETR, TransUNet or Segtran, have their own merits, and should join forces to contribute to the "transformer for segmentation" paradigm. I don't think it helps the community if we fight each other for who was the first that proposed the idea. If you keep an open mindset, I think you are able to find contributions in each work that were absent in the other similar works.
from transunet.
I am glad that my comment here forced you to add the source repository from which you have got the code. So I believe you should educate yourself about adding references to public code you publish ( not me as you suggested ). Again, most of your code is from this repository.
You have cited the paper I mentioned but never mentioned that it proposed a similar architecture before you did, nor you compared against it. Your paper is not considered as concurrent submission to a paper posted months before your paper and submitted to a different conference. You should really admit your mistake, take the high road and post a revised version to ArXiv. You should discuss these matters I brought with your professor and seek advise before submitting this to any actual conference.
from transunet.
Although this work is similar to SETR, the authors do present the results without skip connection (which is essentially SETR?) and show that the U-shape connections is important. @ching-sui1995
@segtran Well said and good luck to your paper review.
from transunet.
Suggesting to acknowledging prior papers is never unprofessional. Whether I bring it up or not, the community can see and understand your contribution and the way you treated that paper. Enough said here.
from transunet.
I think your 2nd suggestion is invalid and truly unprofessional. Please feel free to bring this matter to my advisor or anyone you like.
from transunet.
Related Issues (20)
- Different input size (width x height)
- Reason for [-125, 275] input clipping
- "ZeroDivisionError: integer division or modulo by zero" when vit_patches_size=8 HOT 3
- Need R50+ViT-B_16 rather than R50-ViT-B_16!
- 当我在运行TransUNet-main的train.py时出现错误:KeyError: 'Transformer/encoderblock_0/Multi5HeadDotProductAttention_1/query/kernel is not a file in the archive' 这是在我进行KeyError: 'Transformer/encoderblock_0/MultiHeadDotProductAttention_1/query\\kernel is not a file in the archive'后的更改出现的错误,csdn说这是os.path.join 合并路径的时候出现的问题,更改后仍然出现以上错误 HOT 4
- Even if we fix the seed, the results change for each training.
- asking for you help
- ACDC dataset 100 cases of data HOT 1
- Training for three-channel dataset
- Need R50+ViT-L_16 pretrained model rather than R50+ViT-L_32 HOT 1
- 导入包报错,文件夹重名导致的坑 HOT 2
- Data preprocessing HOT 2
- Training with different image size HOT 1
- Is the Synapse multi-organ segmentation dataset experimental result obtained from is the 20 samples official test set?
- change patch_size during test
- PreActBottleneck
- Training performance issues on small-sized targets
- The issue arises from the absence of the "lists_Synapse" folder.
- ACDC dataset HOT 1
- About the solution of problems like "have 3 channels, but got 1000 channels instead"
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from transunet.