Implements deep learning papers. It contains basic papers such as activations, CNNs to advanced papers such as encoder-decoder etc.
It also contains links to paper from which it is taken.
My Aim is to learn to write research code and easily reproduce it from papers.
Some implementations are mine, most will be taken from various GitHub repos.
My aim here is to learn, reproduce work and understand nuances.
- It presents easily reusable code for Deep Learning Papers.
- It contains
*.pyand easy to reproduce code. - It becomes harder for people to open
*.ipynbfiles and decode stuff and reuse from them. I have created functions and classes which is best practice of writing code. - Code is written in Pytorch so that most community can benefit.
- Contributions for tensorflow code are welcome.
- It is not a repo to showcase my DL skills or show off any implementations.
- Not to beat and achieve benchmarks.
- Not a repo to prove points how one paper is better than other. It depends on context where you apply DL. You won't fit a neural network when decision tree might work.
- Not a repo to showcase how papers keep battling for SOTA. I want to reimplement stuff not create a benchmarking repo.
- Yes as of now. I'm not a great DL researcher / scientist; just a student.
- I would suggest use https://github.com/rwightman/pytorch-image-models this repo for models.
- I have tried to learn from its implmenatations as well.
- Yes you can. It is just reimplementation. But I do not guarentee if that will help you to get SOTA.
- Raise a PR
- Document and comment the code well. Raise a PR.
- No, please do not for this repo is not for such frameworks that have too much built in power.
- But yes please use tensorflow v2 or Pytorch or Pytorch-lightning.