GeoVision-10M

A large-scale geotagged image dataset spanning 195 countries with precise GPS coordinates, timestamps, and comprehensive metadata for training visual geolocation models.

Contribute Data

Dataset Overview

10M+

Geotagged Images

195

Countries Covered

JPEG

High-Quality Format

GeoVision-10M is designed to advance research in visual geolocation, landmark recognition, and urban environment analysis. Each image includes precise GPS coordinates extracted from EXIF data, along with temporal information and camera metadata.

Data Structure

Image Metadata (JSON)

{
  "image_id": "geo_0000001",
  "filename": "geo_0000001.jpg",
  "gps": {
    "latitude": 40.7128,
    "longitude": -74.0060,
    "altitude": 10.5,
    "accuracy": "high"
  },
  "timestamp": "2024-01-15T14:30:00Z",
  "camera": {
    "make": "Canon",
    "model": "EOS R5",
    "focal_length": 24,
    "iso": 100
  },
  "location": {
    "country": "United States",
    "region": "New York",
    "city": "New York City"
  },
  "annotations": {
    "building_type": "commercial",
    "architectural_style": "modern",
    "environment": "urban"
  }
}

Included Fields

GPS coordinates (lat/lng/altitude)
Timestamp and timezone information
Camera make, model, and settings
Geographic location labels
Architectural annotations

Data Quality

GPS accuracy validation
Duplicate detection and removal
Privacy-preserving processing
Balanced geographic distribution
Manual quality verification

Collection Progress

Images Collected2.3M / 10M

We're actively collecting and processing images. Help us reach our goal by contributing your geotagged photos.

Phase 1: Collection

Gathering geotagged images from contributors

In Progress

Phase 2: Processing

Validation, annotation, and quality control

Ongoing

Phase 3: Release

Public dataset release for research

Planned

Help Build GeoVision-10M

Contribute your geotagged photos to help create the world's largest open visual geolocation dataset.

Start Contributing View All Datasets