Product Information
What is Plonk?
Introducing the first generative geolocation method based on diffusion and flow matching! We explore the relationship between visual content and location by treating random positions as images.
New SOTA for visual geolocation on OpenStreetView-5M, YFCC-100M, and iNat-21
Generates global probability density maps and quantifies localizability
Introduces the problem of probabilistic visual geolocation
Our model predicts the location distribution of images from three major visual geolocation datasets: iNat-21 (wildlife), YFCC-100M (natural images), and OSV-5M (street views). It handles varying levels of location ambiguity by outputting appropriate probability distributions.
How to use Plonk?
Plonk is a diffusion and flow-matching generative geolocation method, predicting image location distributions via conditional denoising, achieving SOTA in visual geolocation.
Core Functions of Plonk
AI-Driven
Usage Scenarios of Plonk
- Predict the geographical distribution of wildlife images.
- Predict the geographical distribution of natural images.
- Predict the geographical distribution of street view images.
Common Questions about Plonk
What does Plonk do?
How do I use Plonk?
What are the core features of Plonk?
What are the application scenarios for Plonk?
