Last updated: 22 October 2026. Written by Ritesh Kanjee, director of Augmented AI.
Street Swift Media's delivery bikes carried digital billboards, but GPS only showed where they went, not who saw the ads. In 2022 we deployed an edge AI vision model directly on the bike hardware to count real-time impressions along each route. Mobile billboard reporting moved from estimates to measured data.
In 2022, we partnered with Street Swift Media, an innovative player in the advertising tech industry, to solve a major physical-world tracking problem. By deploying on-bike camera vision, we moved mobile billboard analytics from guesswork to hard data. This edge AI advertising analytics case study breaks down how we built and deployed a production-ready computer vision model directly on delivery-bike hardware.
What problem did Street Swift Media have?
Street Swift Media had a fleet of delivery bikes carrying digital billboards through dense urban environments. While they had GPS tracking to show where the bikes traveled, they had no way to prove how many eyes were actually on the screens. Advertisers demanded hard proof of performance, but the team was left relying on static municipal traffic averages to estimate reach.
Operating in the transit advertising space means facing massive environmental variables. GPS data alone cannot tell you if a bike is riding down an empty alley or stuck in bumper-to-bumper traffic next to hundreds of commuters. To scale their ad network, they needed real-time, verifiable impression counts captured directly from the road.
What did we build?
We designed and deployed an edge AI system that mounts directly onto the delivery bikes. The hardware suite combines a ruggedized, low-power camera with an edge processing unit capable of running local object detection models. Processing video locally on the edge was critical because streaming high-definition video over cellular networks is cost-prohibitive and presents privacy concerns.
The core software runs a highly optimized lightweight convolutional neural network (CNN) trained specifically for vehicle and pedestrian detection. As the bike moves, the camera captures the surrounding traffic environment, registers objects within a specific viewing cone, and calculates estimated impression counts. This allows the system to accurately determine audience density without relying on a persistent cloud connection.
Once the edge device processes the frames, it instantly discards the raw video to maintain privacy compliance. It only transmits lightweight, encrypted JSON metadata containing timestamped count metrics back to the central ad server. This metadata integrates directly with Street Swift Media’s digital board analytics platform, updating their programmatic ad delivery engine in real time.
In short, the system has these parts:
- A rugged, low-power camera and edge processor mounted on each bike.
- A lightweight detection model that counts vehicles and pedestrians in the viewing cone.
- Raw video discarded on the device; only encrypted count data sent to the ad server.
What changed for Street Swift Media?
The deployment of the edge AI system completely transformed how Street Swift Media packages and sells their advertising space. Instead of presenting clients with high-level estimations, they now deliver verified, audit-ready impression reports generated directly from the road. This capability has elevated their inventory from standard out-of-home media to high-precision programmatic advertising.
Here is the before and after in short:
| Area | Before | After |
|---|---|---|
| What clients receive | High-level impression estimates | Verified, audit-ready impression reports |
| Detail | One figure for the campaign | Impressions by hour, route and traffic conditions |
| Pricing | Standard out-of-home rates | Premium pricing tiers |
Advertisers can now see exactly how many impressions their campaigns generate during specific hours, on specific routes, and under varying traffic conditions. If a bike is stuck in traffic next to fifty cars, the system registers the spike in impressions and updates the ledger. This granular level of attribution has unlocked premium pricing tiers and significantly increased overall ad network revenue.
Who is this system for?
This solution is built for operations directors, fleet managers, and out-of-home media executives who need to bridge the gap between physical assets and digital attribution. If you operate transit media, smart city infrastructure, or mobile retail networks, relying on legacy telemetry is no longer enough. You need verifiable proof of presence to secure top-tier brand investments.
Whether you manage digital screens on rideshare fleets, utility vehicles, or delivery networks, this edge AI architecture is highly adaptable. It suits any operator who needs to convert real-world physical activity into structured, privacy-compliant data streams without incurring massive cloud-bandwidth costs. Partnering with a specialized team allows you to build these custom pipelines without stretching your internal software teams.
Common questions
How does the system handle privacy compliance?
The system processes all video feeds directly on the edge device and immediately deletes the footage. No facial recognition or license plate data is ever stored or transmitted, ensuring complete compliance with strict local privacy regulations.
Does edge AI drain the delivery bike's battery?
No, the system is engineered specifically for low-power edge compute platforms. It runs off a highly efficient power management circuit that draws minimal current, preserving the primary battery life of electric delivery bikes.
How does weather or night riding affect detection accuracy?
We optimized the computer vision model to account for diverse lighting and environmental conditions. Using specialized training datasets that include night, rain, and glare scenarios ensures consistent vehicle and pedestrian detection around the clock.
Summary
By moving from physical-world guesses to real-time edge AI validation, Street Swift Media unlocked verified impression tracking for their delivery-bike ad network. This custom edge AI advertising analytics case study demonstrates that complex computer vision models can run efficiently on highly constrained mobile hardware. Operators can now deploy similar systems to prove the value of their physical assets and drive higher ad revenues with confidence.
What is the next step?
If you are looking to deploy intelligent computer vision solutions across your physical fleet, let's build it. You can review the full Case study to see the technical specifications. Reach out to a Business process automation consultant to discuss your integration, or Hire Augmented AI directly to start your build.
