jinxixiang / fista-net

A model-based deep learning network for inverse problem in imaging

Hugo Academic CV Theme

The Hugo Academic Resumé Template empowers you to easily create your job-winning online resumé, showcase your academic publications, and create online courses or knowledge bases to grow your audience.

️Trusted by 250,000+ researchers, educators, and students. Highly customizable via the integrated no-code, Hugo Blox Builder, making every site truly personalized ⭐⭐⭐⭐⭐

Easily write technical content with plain text Markdown, LaTeX math, diagrams, RMarkdown, or Jupyter, and import publications from BibTeX.

Check out the latest demo of what you'll get in less than 10 minutes, or get inspired by our academics and research groups.

The integrated Hugo Blox Builder and CMS makes it easy to create a beautiful website for free. Edit your site in the CMS (or your favorite editor), generate it with Hugo, and deploy with GitHub or Netlify. Customize anything on your site with widgets, light/dark themes, and language packs.

• 👉 Get Started
• 📚 View the documentation
• 💬 Chat with the Hugo Blox Builder community or Hugo community
• 🐦 Twitter: @GetResearchDev @GeorgeCushen #MadeWithHugoBlox
• ⬇️ Automatically import your publications from BibTeX with the Hugo Academic CLI
• 💡 Suggest an improvement
• ⬆️ Updating? View the Update Guide and Release Notes

We ask you, humbly, to support this open source movement

Today we ask you to defend the open source independence of the Hugo Blox Builder and themes 🐧

We're an open source movement that depends on your support to stay online and thriving, but 99.9% of our creators don't give; they simply look the other way.

❤️ Click here to become a GitHub Sponsor, unlocking awesome perks such as exclusive academic templates and widgets

Demo image credits

• Unsplash

Latest news

• 6 Compelling Reasons I Switched from WordPress to Hugo
• The 7 best landing page builders in 2024
• Start a Blog and Make Money in 2024: Here's What You Need to Know
• Hugo vs Quarto: Which One is Better for 2024?
• Easily make an academic CV website to get more cites and grow your audience 🚀

fista-net's Issues

ModuleNotFoundError: No module named 'skimage'

ModuleNotFoundError: No module named 'skimage'

Question about differentiability of computational graph (`torch.sign`)

Heya!

I have a question about the code:

In ISTA and FISTA, you use torch.sign() in the soft threshold step. However, torch.sign() is non differentiable in pytorch, or more specifically, its gradient is always = 0.

I see that the constraint of the symmetric loss may indeed help produce a computational graph, i.e. a gradient != 0. But for gamma=0 (in (13) in DOI: 10.1109/CVPR.2018.00196) the computational graph is broken and the backpropagation will always be zero. Not exactly for the model you have, ISTANet+, but certainly for ISTANet.
For ISTANet++ there would be part of the model that would never get gradients, in particular the part before the soft threshold.

Is this on purpose? I can't see anything in the paper suggesting this is desired, but I may be missing something obvious :)

Thanks for the great work!

if image.dtype.char not in 'df': 报错：AttributeError: 'torch.dtype' object has no attribute 'char'

数据集

你好 不好意思打扰了 因课题需要想复现一下您的代码 请问您可以发布一下数据集吗

关于CT数据集

How to get the EIT Dataset?

Thanks for your excellent work!

And, where can I download the EIT dataset to run your code?
Can you tell me the website?

Thanks!

Training Dataset

The training datasets are not described clearly in main paper. Could you release the training dataset of EMT?

The '.\data' directory and the mat files in it are missing. The './D.mat' file is also missing.

The '.\data' directory and the mat files in it are missing. The './D.mat' file is also missing.

How to get the EIT or CT Dataset?

Thanks for your excellent work! And, where can I download the EIT or CT dataset to run your code?

I would like to carefully inquire whether it is possible to share or make publicly available the dataset you have used. I understand that this may involve sensitive information, and if it does not align with your plans or policies, I completely understand and appreciate your consideration.

If there are alternative ways or resources to access the relevant data, I would also greatly appreciate your guidance.