Auto Bidding: ProovStation provides GPU-driven AI evaluations

Vehicle ratings are boosted by overhauling the AI ​​boosted by the GPU.

ProovStation, a four-year-old startup headquartered in Lyon, France, is undertaking the ambitious pursuit of computer vision to automate vehicle inspections and repair estimates, with the goal of creating ultra-precise AI-powered stations in companies around the world.

It recently launched three state-of-the-art vehicle scanners at the locations of Montaison, Vinicio and Aix-en-Provence, French retail giant Carrefour. ProovStation car scanners have been deployed in Carrefour parking lots for drivers to try out the free service.

Self-service stations are designed for users to provide vehicle and ride information with a value report and repair estimate in less than two minutes. It also enables drivers to get a dealer’s offer to buy their cars as quickly as possible In just seconds – which bodes well for consumers, used car dealers and auctioneers.

Much is run across cameras and sensors, high-resolution graphics, multiple damage detection models, models and analytics to transform damage detection into repair estimates and purchase offers.

“People often ask me how I got so much AI into this, and I tell them I work with NVIDIA Inception,” said Gabriel Tisander, General Manager and Chief Product Officer at ProovStation.

Take advantage of NVIDIA GPUs and Nvidia Metropolis The ProovStation SDKs enable it to scan 5 gigabytes of image and sensor data for each vehicle and simultaneously apply multiple AI vision detection models, among other tasks.

ProovStation used file NVIDIA DeepStream SDK To build an advanced AI insight pipeline and improve inference productivity with AI Triton Inference Servant.

The ProovStation setup enables inference operation for rapid transformations of vehicle analysis in this industry-leading Edge AI Request.

Leadership Progress: Bernard Group Agents

ProovStation posts its stations in a quick clip. This was possible because early-stage founder Gabriel Tesander was linked to an ideal ally in Cedric Bernard, whose family group dealerships and services invested in Bernard Automotive Group for the first time in 2017 to strengthen its own operations.

Groupe Bernard collected vast amounts of image data from its own ProovStation modeling business. Bernard left the family business to join Tissandier as the co-founder and CEO of the startup, joined by co-founder Anton Comizza, and it’s been an intense launch journey ever since.

ProovStation is a member of NVIDIA starta program that accelerates cutting-edge startups through access to hardware and software platforms, technical training, as well as support for the AI ​​ecosystem.

“People often ask me how I got so much AI into this, and I tell them I work with NVIDIA Inception,” said Tesander, General Manager and Product Manager at ProovStation.

Launching artificial intelligence stations across the markets

ProovStation has deployed 35 survey stations in operation so far, and expects to double that number next year. It launched its powerful AI-powered stations in Europe and the United States.

Among the first users to adopt are Groupe Bernard, UK car dealership BCA, OK Mobility car rentals in Spain, and Sixt car rentals in Germany. It also works with undisclosed US automakers and a major online auto seller.

Sixt car rental has installed a terminal at Lyon Saint Exupery Airport to facilitate vehicle pick-up and return operations.

“Sixt really wants to change the car rental experience,” said Tessander.

Creating a “Super AI Factory” for damage datasets

ProovStation has built data science expertise and a team dedicated to handling the many specialized data sets to meet the challenging challenge of damage detection.

“Going from a damage review to a damage assessment can be really difficult at times,” Tesander said.

ProovStation has a team of 10 experts in its Superior AI Factory dedicated to characterizing data with its specialized software. They have processed more than 2 million images with stickers so far, and assigned a classification of more than 100 types of damage and more than 100 types of parts.

“We knew we needed this level of granularity to make it reliable and effective for businesses. Labeling images is critical, especially to us, so we invented some ways to label certain damages.”

Tesander said the data science team members and others have been bred to accelerate AI through courses from NVIDIA Deep Learning Institute.

Introducing data collection with NVIDIA Industrial Edge AI

ProovStation scans a vehicle with 10 different cameras at its station and takes 300 photos – or 5GB of data – to run on its detection models. NVIDIA GPUs enable ProovStation’s 90-second AI inference pipeline to provide damage detection, assessment, localization, measurements, and estimates. The wheels are checked by an electromagnetic frequency device from Michelin Tire Company to estimate wear. It all runs on the NVIDIA edge AI platform.

Using two NVIDIA GPUs in a station allows ProovStation to process all of this in high-resolution image analysis to improve resolution. This data is also transmitted to the cloud so the data science team at ProovStation can use it for further training.

Cameras, lighting, and positioning are big issues. Detection models can get rid of things like glare on cars with shiny windows. ProovStation uses a defect measurement model, which allows it to perform detection while displaying lines on vehicle surfaces, highlighting spots where problems appear in the lines.

It is a difficult problem to solve that leads to job opportunities.

“Every auto industry does car inspections for service — to sell new tires, to fix your car or windshield, it always starts with a car inspection,” Tesander said.

Leave a Comment