The World At CVPR 2025
I wondered where all the paper comes from 😊
As this is my first major large-scale computer science conference, I was really curious where all the people and papers came from! Therefore, I crawled the OpenAccess Archive papers of the main conference (sorry workshops!).
After loading the data, and with a lot of affiliation parsing, LLM-world-knowledge-based affiliation to (city, country) mapping, I had the rough database ready. I think a lot more nice information and statistics can be extracted from it!
For now, I focused on the countries and cities that got their papers published at this year’s conference. Again, I would like to stress that affiliations have been mapped to the city and country via Gemini 2.0 Flash, hence, I rely on its world-view. Please note that city names and locations are obtained from here at simplemaps (Creative Commons Attribution 4.0 License). The country names and borders are from Natural Earth.
Lastly, I matched all data sources resulting in the country data and city data and they are available under the Creative Commons Attribution 4.0 License. Feel free to explore and play around with them, but I cannot guarantee that all entries are perfectly correct (or complete). I tried my best at parsing and filtering the information.
Below you can see an interactive visualization (drag, zoom, and hover) of the data built with 3djs.
Is it perfect? Not at all ^^ but it was still a nice endeavor.
I highlight all cities with papers (aggregated multiple research institutes and universities) and the stats per country (the color grading should be a clear indicator ^^).
The papers per country tells us how often a country (with at least one affiliation residing in there) has been part of a publication.
Let’s see else what you can find there!
Downloads
- CityData (csv file)
- CountryData (csv file)