Computer Vision / Video Analytics

Metropolis Spotlight: Sighthound Enhances Traffic Safety with NVIDIA GPU-Accelerated AI Technologies

JAKARTA, Indonesia - December 17, 2019: Tilted top down horizontal view of a fountain in the middle of a roundabout in Jakarta city at sunny morning.

NVIDIA Metropolis partner Sighthound—formerly Boulder AI—is helping cities improve traffic management and pedestrian safety with software and hardware solutions to bring cloud-native solutions for edge data intelligence.

To design efficient, equitable, and sustainable infrastructure, city planners rely on accurate roadway usage data. Sighthound builds edge-enabled, roadway-focused products that deliver data to cities for applications that protect pedestrians in intersections, quantify parking occupancy, and analyze highway utilization.

A Competitive Advantage for Livability and Safety 

Sighthound’s primary goal in the highly competitive AI landscape focuses on using the most advanced tools along with a full-stack approach to achieve faster time-to-market compute-optimized solutions. They leverage the powerful capabilities of the embedded NVIDIA® Jetson™ edge AI platform—which provides GPU-accelerated computing in a compact and energy-efficient module—that powers DNN cameras and nodes.

Combining its industry-leading proprietary developed traffic data set with the pretrained TrafficCamNet model from NGC, and the NVIDIA TAO Toolkit, Sighthound fast tracked the AI development pipeline. It also refined the performance and accuracy of the platform. Cutting development time to just a few weeks, it delivered a traffic safety solution to the City of Denver that automatically detects, counts, and measures speed and tracks pedestrian, bicycle, and vehicles occupancy.

“The pretrained NVIDIA models helped us to get great accuracy for our traffic models and go to market faster. It allowed us to focus on building the surrounding technology and extract the real-time data, which creates real value for our customers and ultimately for the citizens who live in these smarter cities,” said Darren Odom, CTO of Sighthound, former CEO of Boulder AI.

As one of America’s fastest-growing metropolitan areas, the City of Denver faces the challenge of quickly and safely transporting people on aging and constrained infrastructure. The city is deploying Sighthound’s AI-enabled video detection and analytics application to make roads, intersections, and crosswalks safer and smarter. The sensor platform meets Denver’s traffic demands in real-time, allowing the city to react quickly and manage traffic during the most congested scenarios. 

Figure 1. Sighthound traffic AI application in action.

City of Denver Use Cases


Automatic pedestrian detection: Detects pedestrians and provides accurate count, speed, and direction for pedestrians. 

Intent to cross: Predicts when pedestrians intend to cross an intersection, ensuring the timing of traffic signals is aligned with the pedestrian flow. 

Touchless crosswalks: Detects when pedestrians are at or absent from crosswalks, allowing contactless crossing and optimizing traffic flow.

Near miss detection: Captures near-miss events between vehicles, pedestrians, and bicycles, and uses the data to proactively prevent accidents.

Intersection safety index: Tracks statistics like the number of red light violations, speeding vehicles, and illegal turns at each intersection. 

Multimodal data collection: Counts vehicles and measures their speed, turning lane movements and occupancy. 

Crowd detection: Detects crowds at intersections and notifies traffic signal systems in real time to extend their crossing phases for larger crowds.

Check out pretrained models on NGC to jumpstart your AI development.

Discuss (0)

Tags