Hever Amir is the co-founder and CEO of UVeye, a cutting-edge computer vision technology startup. A few years ago, while driving into a government facility, he was stopped by a security guard who immediately checked the bottom of his car. This experience made him realize that traditional security inspection methods were inefficient and unreliable.
"When I got out, I asked the guard what he was looking for," Hever recalled. "He was very honest and said he was checking for potential threats, but nothing was found. That moment made me realize that visual inspections alone aren't effective at detecting real dangers."
Inspired by this realization, Hever formed a team to explore solutions. In 2016, they launched UVeye in New York and developed a vehicle floor detection system powered by deep learning, designed to enhance safety and accuracy.
UVeye focuses on monitoring the undercarriage of vehicles and identifying changes that occur over time. Hever explained, "It's not easy to spot anomalies in a car’s chassis because there are no clear standards to define what’s normal. Plus, threats often hide in plain sight."
The company quickly realized that relying solely on manufacturer-provided chassis data wasn’t enough. After thousands of miles, the structure can change significantly. To address this, UVeye developed unsupervised learning algorithms capable of detecting even subtle threats, regardless of how deep they are embedded.
To train their models, UVeye leased hundreds of cars in various conditions, scanned their undercarriages, and created 2D images and 3D models. These datasets were fed into their deep learning system, which maps all components and analyzes them individually to identify any irregularities.
This system can detect even small foreign objects, like a USB drive, and determine whether an abnormal protrusion is just snow or something more dangerous.
UVeye uses high-performance workstations equipped with multiple NVIDIA GPUs for training. When needed, they also leverage cloud-based GPU resources from Amazon Web Services and Microsoft Azure to accelerate processing.
Hever emphasized that the use of GPUs and CUDA parallel computing has greatly improved the speed and efficiency of their development process, enabling faster and more accurate results.
UVeye’s first product line allows users to automatically scan, detect, and identify various chassis issues, including foreign objects or structural changes. The system is already deployed in over 30 locations worldwide, scanning vehicles as they pass by. This continuous data collection ensures the system remains reliable and effective.
Hever noted, "GPUs make fast detection possible. Our machine learning algorithm can analyze a moving car and detect anomalies in just three seconds."
Beyond the undercarriage, UVeye’s system can inspect other critical parts of the vehicle, offering a full-range inspection. Hever said, "Our 360-degree system can detect cracks, wear, and other damages across the entire vehicle."
From car sales and leasing to fleet management and maintenance, ensuring safe and reliable operations involves many aspects. Hever sees endless opportunities for his "detection as a service" model, which aims to revolutionize how people and institutions inspect vehicles.
"We will change the way people detect cars," Hever concluded. "Our goal is to make safety smarter, faster, and more accessible."
understanding mining,outdoor mining,Mining Fiber Cable
Guangzhou Jiqian Fiber Optic Cable Co.,ltd , https://www.jqopticcable.com