- Data & AI Solutions
- Off-the-Shelf Datasets
- U.S. traffic sign character recognition dataset
- images
U.S. traffic sign character recognition dataset
Specialized dataset designed for training and validating highly accurate traffic sign detection and character recognition systems for autonomous vehicles operating in the U.S. Comprises over 10,000 images extracted from more than 15 hours of dash-cam footage recorded in the San Francisco Bay Area.

Specifications
- Modalities
- 2D images
- Location
- Mountain View and South Bay areas (San Francisco)
- Licensable
- Yes
- Sensor setup
- 1x dash-camera
- Data volume
- 15 hours 52 mins of recording @30 FPS
- Total images
- 10,868 annotated images
- Year of collection
- 2024 and 2025
- Time of day
- Afternoon, evening, morning, noon, night, dusk, dawn
- Licensable
- Yes
- Environment distribution
- Sunny: 38.32%; partly cloudy: 33.50%; mostly sunny: 16.03%; cloudy: 12.15%
Accelerate model development & training processes
Diverse sign coverage
Strategically captured to include a rich mix of regulatory, warning and informational signs, ensuring broad applicability and robustness in real-world deployments.
Optimized driving route
Route planning that ensures high exposure to diverse road signs while covering varied terrain and infrastructure in San Francisco.
Environment and weather diversity
Captured across varied times of day, from dawn to night and in real-world weather conditions, this dataset helps improve model generalization and resilience.

Recommended datasets
See allExplore our success stories
Curating high-quality data for the training and validation of ADAS and AV models
12TBdata captured daily
7500kmapproximate total distance covered
Curating high-quality data for the training and validation of ADAS and AV models
12TBdata captured daily
7500kmapproximate total distance covered
Upgrade your ADAS and AV models
Partner with our AI experts to advance your perception and autonomy models.
Explore our custom AI solutions
