Agentic AI model for visual geolocation that combines image analysis with live web search to pinpoint where a photo was taken. It uses an image-zoom tool to inspect details and a web-search tool to validate hypotheses trained on GeoBench, a new benchmark of global photos, panoramas, and satellite imagery. Outperforms open-source models and rivals GPT-5/Gemini in accuracy.
https://ekonwang.github.io/geo-vista/
You can found all the other resources here, for free:
Hi, I'm Mario Santella – Security Researcher focused in OSINT and IT guy
With a solid background in IT and software development since 2012, I’ve gradually shifted my focus towards investigative cybersecurity, OSINT (Open Source Intelligence),
and digital threat analysis. My work sits at the intersection of security research, data-driven investigations, and strategic intelligence.
I help individuals, journalists, and organizations uncover digital traces, map risks, and stay ahead of threats in the open web and beyond.
I operate remotely from Italy and collaborate on research, recon, and intelligence-driven projects worldwide.
Linkedin resume in
Github profile