Adding a document with papers only

[mihaibujanca]

I was thinking a list containing the latest papers in computer vision similar to https://github.com/kjw0612/awesome-deep-vision would be great, maybe in a separate .md file linked in the main README.md.

I would be interested in participating actively in this.

[pmoulon]

In the same spirit of @mihaibujanca I made a list of paper/ressources about 3D reconstruction from images https://github.com/openMVG/awesome_3DReconstruction_list
@jbhuang0604 I think it would be valuable to add it somehow in your list. FYI, I list your list in the top of my document.

[mihaibujanca]

@pmoulon @jbhuang0604 I think it'd be very useful to decide on a way to structure the content, it's probably better than me creating a whole new repository - I have my own list offline and I'm adding a few papers every week, so I might as well help with a public list and get other people to review what I'm listing

[jbhuang0604]

@mihaibujanca @pmoulon It's a great idea! However, I am not sure how to organize such a list of papers due to the diversity of the topics. Any ideas?

@pmoulon This a great list! I added the link to the multi-view vision.

[mihaibujanca]

My suggestion would be either

Have a directory containing the files, each of them on a particular topic, or
Have a document with papers, with a table of contents listing each topic, and the document structured as

• Topic
  • Year
    • (optional) Conference that took place that year
      • List of papers in alphabetical order

Images related to the topic, similar to the awesome deep vision repo would be useful as well

[mihaibujanca]

The format I would suggest for listing papers is
Paper Title. Authors, Conference Paper, Code, Web

e.g
Parallel Tracking and Mapping for Small AR Workspaces. Georg Klein, David Murray, ISMAR 2007, Paper, Code, Web

[pmoulon]

Just a short note:

• On my repository the paper name is already a link to the paper, it allows to compact a bit the things.

[mihaibujanca]

That's a good way of doing it as well

[mihaibujanca]

I think the main problem here is deciding what a "topic" is and how fine-grained do we want the topics to be.
We could possibly have separate files for wider issues. For example for 3D reconstruction, which has a lot of sub-topics like SLAM, SfM etc we could add Pierre's list as a sub-repository. Similarly we could have stuff like 'scene understanding' which has normals estimation, boundary detection, layout estimation and so on.

The issue with having separate files might be that we will end up with the same resources in multiple files, which is easier to avoid when there's only one file. On the other hand some of the papers do belong to multiple categories