Quividi Adds AI-Driven Motor Vehicle Counting, Classification, To DOOH Audience Measurement Toolset
September 22, 2021 by Dave Haynes
The longtime audience measurement firm Quividi, which has been doing computer vision long before that name grew common, has added vehicle detection and classification to its feature set – with the intent of helping more accurately characterize motorist audiences for roadside digital OOH networks.
Interestingly, and I am not entirely sure why, the French firm is using Intel’s open source computer vision library – OpenVINO – as the main tool for this new feature. I say interesting because Quividi has been around for way more than a decade and has long had its own video analytics platform.
It may have been easier, faster and less costly in resources to adapt code already written than to produce it from the existing code-base. The Quividi PR suggests this will in parallel to Quividi’s established audience measurement tools.
Quividi says the “solution counts, in real-time, the numbers of vehicles passing next to a point of interest and breaks it down by type (car, bus, van, etc.) to generate live traffic and impression data, along with presence time. Vehicles are detected from 200 meters away with a 98% accuracy during daytime and 92% during nighttime. Impression data generated from the vehicle count is calculated based on the official average number of passengers per vehicle in a given market (i.e. 1.9 persons per car, 7 persons per bus, etc.).”
The feature is included into AMP Outdoor, Quividi’s new platform to measure the performance of downtown DOOH assets – e.g. roadside billboards, city kiosks, EV charging stations, shop window screens – with their mix of pedestrians and vehicles audiences. The vehicle detection solution runs simultaneously to Quividi’s existing attention and footfall solutions, while the live audience data is aggregated into a unified web dashboard.
AMP Outdoor is already being used by a selection of Quividi partners across the world.
“Quividi is proud to provide the most comprehensive and accurate solutions to measure consumer engagement with DOOH, now on all types of venues,” says Olivier Duizabo, President of Quividi. “With the introduction of our new vehicle detection feature, we provide a complete audience measurement suite that powers smart DOOH campaigns outside, inside, and in-store.”
One of the handful of companies that Quividi competes with for Digital OOH network operators is Miami-based AdMobilize, which released similar tech about four years ago – called Vehicle Recognition Engine.
Hi Dave, quick reply on your question re: OpenVino.
Over the years, Quividi has developed a very effective hybrid approach to computer vision: the core algorithms powering our audience measurement solution are developed in-house with a focus on optimization, so that our software can run even on very low-end platforms.
When additional computational resources are available (as is generally the case when monitoring vehicular traffic), our software will make use of more sophisticated inference models.
Intel’s OpenVINO is an extremely efficient general-purpose inference library; it provides state-of-the-art performance on sufficiently capable hardware and it fits seamlessly in our modular processing architecture.
In the past couple of years we have worked closely with Intel to fine-tune OpenVINO for 24/7 operation at the edge and today Quividi’s network represents the largest deployment of OpenVINO instances to date.
Olivier