AI, 3D, and 360 degree vehicle imaging and inspection (Manheim and wholesale digitization)

Overview

Fyusion is a computer vision and immersive imaging company acquired by Cox to deepen wholesale and retail condition capture—think walkarounds, damage, 360s, and faster digital remarketing for Manheim and partners.

Notes

Cox and Manheim use Fyusion to image at scale so remote buyers and upstream channels can make decisions on condition without being in the lane. The IP is in 3D, CV, and mobile-first capture.

Research https://www.fyusion.com. Write 2500+ word expanded editorial.

Fyusion Expanded Editorial: AI-Powered 3D Vehicle Imaging and the Digitization of Wholesale Automotive

Introduction

Fyusion is a San Francisco-based visual intelligence company that sits at the intersection of 3D imaging, computer vision, and artificial intelligence. Acquired by Cox Automotive in January 2021, Fyusion provides the core imaging and condition-capture technology behind Manheim's digital wholesale marketplace — the world's largest. Its technology enables anyone with a smartphone to capture a comprehensive, 360-degree 3D model of a vehicle in minutes, then have AI analyze that model for damage, condition, and features with accuracy that rivals in-person inspections.

Fyusion's domain is wholesale vehicle digitization at industrial scale. With over 150 patents, 40+ PhD-level professionals on its workforce of roughly 60, and an open-source machine learning library (FyuseNet) released to the developer community, the company punches well above its size. Its technology powers everything from Manheim Express (the mobile app dealers use to create wholesale listings in minutes) to fixed imaging tunnels at physical auction locations across the country. This editorial covers Fyusion's founding, technology, acquisition, product ecosystem, and strategic significance within Cox Automotive.

Founding and Early History

Fyusion was founded in 2014 (some sources cite 2013) by Radu B. Rusu, Stefan J.J. Holzer, Stephen D. Miller, and Pantelis Kalogiros. All four co-founders had deep roots in computer vision and artificial intelligence research. Rusu, the CEO and co-founder, is a world-renowned expert in 3D data processing and was named one of Goldman Sachs' Top 100 Most Intriguing Entrepreneurs in 2016. He also received the IEEE RAS Early Career Award in 2013 for contributions to the field of 3D computer vision.

The company's origin story is rooted in a specific insight: visual intelligence was being held back by a reliance on 2D data — traditional photographs and video. The co-founders believed that developing 3D file formats, combined with software that could assess the world in three dimensions (the way the human eye does), would be transformative. They developed and patented a new file format called the .fyuse, a lightweight 3D capture that could be created using an ordinary smartphone. This was followed by the AI-Based Lightfield Information Suite (ALIS), a scalable AI platform that analyzes 3D images in real time.

Interestingly, Fyusion's technology first made its mark in the consumer social media space. The company's early 3D imaging technology attracted over 100 million users through consumer-facing apps before the company pivoted to enterprise applications — specifically automotive. This consumer proving ground gave Fyusion an enormous dataset and years of real-world refinement on capture quality, file size optimization, and cross-device compatibility before it ever touched a wholesale vehicle.

The Core Technology: ALIS and the .fyuse Format

Fyusion's technology stack is built around three pillars: Capture, Engine, and Viewer — collectively branded as ALIS (AI-Based Lightfield Information Suite).

Capture

The Capture module enables untrained users with common smartphones to create high-resolution, 3D vehicle imagery. The process is guided: a user walks around a vehicle one or two times, and real-time feedback on the phone screen helps them stay on track. The software works in portrait and landscape orientations and is optimized for low- or no-connectivity environments — the patented .fyuse can be created entirely on the device without an internet connection, which is critical for use in underground parking garages, remote lots, or areas with poor cell reception.

Capture supports three modes:

• Walkaround (most common): The vehicle stays static while the user walks around it with any smart device.
• Turntable: The vehicle rotates on a turntable with a static camera positioned at distance.
• Drive-Through (Fixed Imaging Tunnels): Dedicated units with multiple light sources and cameras capture vehicles as they drive slowly through — used at high-volume Manheim auction locations.

Fyusion is compatible with iOS and Android smartphones, as well as professional cameras and Ricoh Theta devices for interior 360-degree captures.

Engine

The Engine is where ALIS does its heavy lifting. It creates the 3D .fyuse model and simultaneously extracts actionable information from the visual data. Patented technology ensures smooth movement within the interactive 360-degree environment, reduces or blurs background distractions, and can automatically tag features or damage on the vehicle.

The key differentiator of the Engine is its 3D-native approach. Most AI-driven vehicle inspection solutions work with traditional 2D photos, which means they can only see what a flat image shows — leading to inconsistent results. Fyusion's Engine analyzes the vehicle in full 3D space, seeing every angle much like the human eye does. This multi-angle capability enables the AI to know exactly where on the vehicle damage is located, what type of damage it is (scratch, dent, crack, etc.), and how severe. The Engine can even compensate for less-than-ideal conditions — it knows if a vehicle is dirty or wet and adjusts its analysis accordingly.

Viewer

ALIS outputs 3D images as .fyuse files, which provide an immersive 360-degree experience with fast load times. The .fyuse format is notably more efficient than traditional formats — at equivalent quality, a .fyuse file averages 54kb compared to 240kb for a .jpg and 401kb for a .png, all while being fully interactive and enabling multi-faceted interactivity. The Fyusion Viewer can be integrated with most content management systems, websites, and applications.

The Viewer can automatically generate a gallery of 2D photos extracted from the 3D capture (standard multi-angle merchandising photos), highlight damages or features with color coding, create 3D closeups that pop open within the image, and steer end users to specific areas of interest.

Products and Solutions

Fyusion's technology surfaces through several distinct products and integrated solutions:

Inspect3D

Inspect3D is Fyusion's AI-driven damage analysis tool for quick and effective condition reports. It was designed specifically for the inspection and remarketing workflow. After a walkaround capture, the system produces a comprehensive condition report delivered to the user's smart device in minutes. The report includes:

• An interactive 360-degree view of the vehicle
• A top-down rendering highlighting all detected damage
• The ability to zoom in on every ding and scratch
• A full breakdown with 2D multi-angle images of all detected damage
• 3D tagging of damages (automatic or user-added) to provide deeper detail

Inspect3D can be augmented with remote human auditing for the highest possible accuracy, creating a human-plus-machine feedback loop. The product is SOC II compliant, fast, and secure.

Auto3D

Auto3D is Fyusion's vehicle merchandising and imaging solution focused on creating beautiful interactive 3D listings for dealers. It enables dealers to shoot, edit, and publish high-quality 3D content from any smartphone, with automatic generation of a 2D photo suite. Key features include:

• A/V tags (audio and video annotations on vehicle features)
• 3D tags that pop open within the 3D environment
• Interior 360-degree captures with Ricoh Theta integration
• Photo playlists (automatic 2D photo sets from a single 3D capture)
• Turntable and freestyle capture modes

Studies cited by Fyusion show greater than 10% increase in engagement with Auto3D's immersive images compared to standard photography.

Next Inspect

Next Inspect (launched in 2023) is a Cox Automotive solution powered by Fyusion's guided vehicle image capture featuring 360-degree multi-angle view imaging. It targets consumers, dealers, and fleet operators who need to inspect vehicles remotely. It's an end-to-end solution for gathering condition information anytime, anywhere, by anyone.

Manheim Express

The Manheim Express mobile app was the original integration point between Fyusion and Cox Automotive, dating back to 2018. It allows dealers to walk a car, create a wholesale listing in minutes with the industry's first 360-degree and movable images, and publish it to the Manheim Marketplace. Audio and video tags were later added to further highlight vehicle condition.

Fixed Imaging Tunnels

Manheim has deployed Fyusion-powered fixed imaging tunnels at its auction locations. These drive-through capture units can process vehicles at high volume, producing consistent, high-quality 3D captures and condition data for every vehicle entering the auction pipeline.

FyuseNet (Open Source)

In March 2023, Fyusion open-sourced FyuseNet, an OpenGL(ES)-based neural network inference library that enables running machine learning models on almost any mobile device — not just flagship phones. Created internally in 2016, FyuseNet is written in portable C++ and runs on desktop, edge, and mobile platforms. It is available on GitHub under the MIT License. The library fuses operations into single shader layers and leverages GPU raster operation processors for essentially free arithmetic. This democratization of mobile ML reflects Fyusion's engineering philosophy and has implications beyond automotive.

The Cox Automotive Acquisition

On January 5, 2021, Cox Automotive announced its acquisition of Fyusion. The deal came after years of collaboration dating back to 2018, when Fyusion first provided imaging capabilities for Manheim Express. By the time of acquisition, Fyusion was already deeply embedded in Manheim's digital transformation.

Steve Rowley, then President of Cox Automotive (now CEO), framed the acquisition around scale and speed: "Anywhere a car can be imaged, we believe the new technologies that Cox Automotive and Fyusion create together will translate to big benefits for our clients across the auto industry."

The acquisition was strategically timed. The COVID-19 pandemic had massively accelerated digital adoption across the Manheim Marketplace. At the time of the acquisition announcement, 83% of Manheim's inventory was already being sold to digital buyers — a number that had surged during the pandemic. Manheim President Grace Huang stated: "Together with Fyusion, we plan to boost client confidence and trust by making the digital buying experience as good as seeing it in person."

Fyusion's co-founder and CEO Radu Rusu emphasized the opportunity: "Combining our industry-leading imaging and AI expertise with Cox Automotive's reach across the industry will bring game-changing innovations to dealers, commercial clients and consumers."

Post-acquisition, Fyusion operates as a distinct brand within Cox Automotive's portfolio. The fyusion.com website now redirects to coxautoinc.com, though archived captures show Fyusion maintained its own brand identity and site until at least early 2025.

Strategic Impact and Industry Significance

Fyusion's technology sits at the center of a fundamental shift in the wholesale automotive industry: the transition from physical lane auctions to digital marketplaces. In the traditional model, buyers would walk auction lanes, inspect vehicles in person, and bid live. Condition reports were subjective, handwritten, and inconsistent. Fyusion aims to solve all of these problems at once.

Consistency: AI doesn't have bad days and never needs coffee. Fyusion's AI produces the same quality output every time, removing the human subjectivity that has plagued condition reporting since the first auto auction in 1945 (Manheim, Pennsylvania).

Objectivity: By analyzing vehicles in full 3D space from millions of data points, Fyusion can detect damage the naked eye might miss. The system catches things that even experienced inspectors overlook.

Speed: A walkaround capture takes minutes. Processing takes minutes more. The entire workflow from capture to published listing can happen in under 10 minutes.

Scale: Manheim offers nearly 7 million used vehicles annually. Fyusion's technology is built to process that volume, whether through mobile walkarounds or fixed imaging tunnels at physical locations.

Remote Confidence: The end goal is making remote buyers as confident in a vehicle's condition as if they had inspected it in person. This unlocks a larger buyer pool, reduces arbitration, and increases the velocity of wholesale transactions.

AI in Auto Remarketing: The Road Ahead

An insight article from Cox Automotive (January 2023) laid out the vision: AI in auto remarketing isn't science fiction — it's already in production. Fyusion's 3D imaging technology is being used to look at vehicles from every angle, and Manheim has spent years training a next-generation automated damage detection model with "phenomenal results" in testing. The next evolution of condition reports will include color-coded interactive images that provide a deeper understanding of a vehicle's condition, moving far beyond static photos.

Since Manheim and Fyusion have more vehicle data than any competitor in the industry, their AI model has a structural advantage: more training data leads to better predictions. This creates a widening moat as the system improves over time.

Use Cases Beyond Wholesale Auctions

While Manheim remains the flagship deployment, Fyusion's technology has broader applications across the automotive ecosystem:

Dealer Merchandising (New and Pre-Owned): Higher quality imaging and automatic feature/damage tagging for online listings, driving higher engagement and faster sales.

Dealer Appraisals (Trade-In and CPO): Reducing appraisal time while increasing valuation confidence. Fyusion documents every detail, supporting equitable trade-in valuations and certified pre-owned certification.

Chain of Custody: As vehicles move from OEM to dealer, from customer to service provider, or through transportation, Fyusion's condition reports provide documented proof of condition at each handoff point, reducing disputes.

Third-Party and Private Sales: Mobile-first capture and condition reporting put professional-grade imaging and inspection in the hands of private sellers and buyers.

Fleet Management: Servicing, appraisals, and condition documentation at scale for commercial fleet operators.

Collision Repair: AI-based collision reports that document pre-existing damage before repair work begins, reducing liability and improving estimates.

Open Source and Engineering Culture

Fyusion's decision to open-source FyuseNet under the MIT License in 2023 reflects the company's engineering-driven culture. Stephen Miller, Fyusion's Co-Founder and Chief Product Officer, stated: "We know how powerful FyuseNet is, as we've used it for years in our own work, and we believe that this can be the one beautiful library that everyone uses to accelerate advancements in machine learning."

The library addresses a real pain point: mobile ML model deployment is notoriously fragmented across operating systems and device generations, meaning powerful AI features are often reserved for expensive flagship phones. FyuseNet's lightweight implementation works on the majority of devices regardless of manufacturer or processor, even without network connectivity.

Competition and Market Position

Fyusion operates in a competitive landscape that includes traditional condition report providers, mobile imaging startups, and emerging AI inspection companies. Its advantages include:

• Cox Automotive distribution: Access to Manheim's 7 million vehicles/year and 40,000+ dealer clients
• Data moat: Years of 3D vehicle data for training AI models
• Patent portfolio: 150+ patents protecting the .fyuse format, ALIS architecture, and capture methodology
• Technical talent: 40+ PhD-level researchers and engineers across 16 countries
• Proven scale: Technology already processing millions of vehicles annually in production

Conclusion

Fyusion represents a case study in how specialized computer vision IP, when paired with the right distribution channel, can transform an entire industry vertical. The company's journey from a consumer 3D imaging app with 100 million users to the backbone of the world's largest wholesale vehicle marketplace is a testament to the power of foundational technology — 3D capture, AI analysis, and lightweight file formats — finding its ideal market fit.

For Cox Automotive, the Fyusion acquisition was not just about adding a technology bolt-on. It was about owning the digital inspection layer of the entire wholesale ecosystem. As the industry continues to move away from physical lanes and toward always-on digital marketplaces, Fyusion's technology becomes more critical: it is the bridge between the physical vehicle and the digital representation that buyers trust to make multi-thousand-dollar decisions without ever touching the car.

The Fyusion story is still being written — automated damage detection continues to improve, fixed imaging tunnels are rolling out to more locations, and the underlying ALIS platform continues to evolve. But the thesis is proven: 3D + AI + mobile-first capture = the new standard for vehicle condition information at scale.