Amazing work by @nicolasdufour.bsky.social 👏
I personally believe this is huge! Even from university regression was an ill-defined task: you optimize MSE and you end up regressing a mid average w/o meaning. Instead what you need to regress is a distribution. This paper solves that and is very fast.
🌍 Guessing where an image was taken is a hard, and often ambiguous problem. Introducing diffusion-based geolocation—we predict global locations by refining random guesses into trajectories across the Earth's surface!
🗺️ Paper, code, and demo: nicolas-dufour.github.io/plonk