geoinfer-image-geolocation_skill

Overview

This skill performs AI-driven image geolocation without relying on GPS, EXIF, or other metadata. It analyzes visual features like architecture, terrain, vegetation, and signage to predict where a photo was taken, typically to city level. The tool is designed for OSINT, digital forensics, investigative journalism, and security workflows. It requires the GEOINFER_API_KEY environment variable to operate.

How this skill works

You provide a local image file and select a model; the skill sends the image to the GeoInfer API and returns a ranked list of predicted locations with coordinates and confidence scores. Models are specialized (global, car, property, accuracy) and consume credits per prediction. Scripts are provided to list available models, run predictions, and check credit balance. Images must be local and under the service size limit.

When to use it

• Investigating the likely location of an image when metadata is missing or stripped
• Rapidly triaging large collections of photos to prioritize geolocated leads
• Verifying or cross-checking reported locations in news or social media posts
• Security and threat assessments that require location context from imagery
• Forensic workflows where image provenance needs corroboration

Best practices

• Set GEOINFER_API_KEY as an environment variable before running scripts
• Run scripts/models.sh first to pick the most appropriate model for your use case
• Use top_n=1 for fast single predictions; use top_n=5+ for comparative confidence analysis
• Check your credit balance before batch processing to avoid interruptions
• Download remote images locally before passing them to the prediction script

Example use cases

• OSINT analyst geolocates a crowd photo to identify the city and nearby landmarks
• Journalist verifies the origin of a user-submitted image for a news story
• Security team assesses potential threat imagery by narrowing down likely locations
• Forensic investigator triages a set of images to find ones with high location confidence
• Real estate researcher uses the property model to approximate where a listing photo was taken