Deploying machine vision is no longer just about installing cameras and writing inspection code; it’s about creating reliable, scalable systems that deliver measurable results in real production environments. In this session, Vista Solutions will share the key lessons learned from decades of deploying advanced vision systems, including AI-driven inspection platforms, across industries such as automotive, medical devices, and consumer goods.

We will explore the foundational framework of any successful deployment - Image, Analyze, Automate - and explain why focusing exclusively on analysis can derail timelines and outcomes. Attendees will gain practical insight into how to manage expectations during a project, what level of ownership is required from end-users (especially for AI vision), and the most common bottlenecks that can hinder a project’s success. We’ll also outline the capabilities that manufacturers and operations teams should develop internally to accelerate their success with machine vision, ensuring projects don’t just go live but deliver sustained value.

Peter Denzinger

Vice President of Engineering
Vista Solutions

In this session I will dive into how to make AI integration easy with machine vision deployments. I will discuss real world examples on how to setup an AI system for real world deployment success. I will walk through the steps needed to setup, image, program and deploy. I have lots of great examples on ease of use and how to continue to be successful over time.

Eric Hershberger

Principal Applications Engineer
Cognex

Deep learning has improved stereo vision in challenging real-world conditions, but it does not eliminate the fundamental limits imposed by system design. This presentation examines where learning-based stereo delivers meaningful practical gains and where performance remains limited by optics and geometry.

We evaluate both hybrid and fully learning-based stereo pipelines across a range of operating conditions. Compared with traditional stereo methods, learning-based approaches demonstrate improved performance in challenging scenarios, such as low-texture regions, partial occlusions, and certain transparent or reflective objects.

By correlating disparity accuracy and depth coverage with key system parameters, we identify clear operating conditions and performance limits. Factors such as optical quality, baseline geometry, and calibration retention impose constraints that cannot be overcome by learning alone.

Attendees will gain actionable insights into designing more reliable stereo vision systems, along with practical guidance to avoid common pitfalls in real-world deployments.

Stephen Se

Senior Engineering Manager, Depth Sensing
Teledyne

SO/IEC 42001 is the first global standard for Artificial Intelligence Management Systems (AIMS), providing a framework to ensure AI technologies are ethical, transparent, reliable, and aligned with organizational and societal values. This presentation explores the significance of ISO 42001, its key components, and its role in shaping the future of AI governance. The content will delve into the standard's principles, including ethical considerations, risk management, and performance metrics, and discuss how organizations can implement it to build trust and compliance in their AI initiatives.

Zeeba Mercer

Accreditation & Technical Manager
TÜV Rheinland of North America Inc.

Paul Thomas

Director of Machine Vision and Applied AI in Global Engineering
P&G

Most of the world's manufacturing capacity already exists. The hardest automation problem isn't building greenfield plants; it's automating brownfield ones, where existing layouts, mixed-model production, and accumulated part variation make traditional fixturing-heavy approaches slow and expensive.

This session looks at how AI-driven 3D vision, mounted on the robot itself, is reshaping what brownfield automation can achieve. Stellantis will share lessons from deploying this approach across multiple stations: how robot-mounted vision lets cells adapt to part variation in real time, how it unlocks automation in operations that previously stayed manual, and what it changes about the build-vs-buy and retrofit-vs-rebuild decisions plant teams face. Inbolt will cover the technical context, what the vision system needs to deliver, how deployments are scoped on existing cells, and where the limits are.

The session is non-commercial and focused on practical lessons for engineers and operations leaders evaluating AI vision in their own plants.

Albane Dersy

COO & Co-founder
Inbolt (with Stellantis)

Carl Standertskjold

Head of Innovation - North America Manufacturing
Stellantis

Annie Lu, CEO & Co-Founder of Laminar, has spent her career inside the facilities most technology companies overlook – legacy process manufacturing plants running on equipment and workflows built decades ago. With 100+ deployments across six continents, she has seen firsthand how food & beverage operations bleed margin through wasted water, chemicals, energy, and time – not from carelessness, but from systems that were never designed to learn or adapt. The business case for fixing that is now undeniable: facilities deploying Laminar’s process-aware AI are seeing an average of a 20% reduction in water and chemical consumption and 15% uptime, achieving ROI in under a year. With productivity and sustainability both strived for, this session is the unfiltered vision on how those numbers get built, and what it takes to get there.

The pressures bearing down on process manufacturing today make the cost of standing still harder to justify every year. Raw materials costs are rising, SKU complexity is exploding, and decades of tribal knowledge are walking out the door faster than it can be captured. Yet the “we’ve always done it this way” mindset remains one of the most powerful forces in legacy manufacturing – and understandably so, when the stakes are product quality, food safety, and margins that leave no room for mistakes. Annie will be direct about the conversations that happen on the factory floor and why the zero-ask principle – no infrastructure overhaul, no data science team, no disruption to existing operations – is what ultimately unlocks the business case in facilities that have seen too many technology promises fall flat.

Through real case studies with some of the world’s largest manufacturers like Coca-Cola, Unilever, and AB InBev, attendees will understand how process-aware AI operates in three steps: sense real-time fluid conditions, reason the optimal action, and act on it. Process lines that once depended on experienced operators making judgement calls and timer-based recipes now autonomously adjust cleaning steps, chemical concentrations, temperatures, and changeover transitions – improving yield and quality consistency every shift while compounding savings over time.

What made global scale possible was a commitment to first-principles thinking and deep collaboration with the people closest to the process. She’ll walk through what the world’s largest manufacturers asked for, what it took to earn their trust, and what it means to guarantee quality and safety on every single shift and changeover. Annie will explain where the savings come from, how they are measured, and how lessons have shaped the technology and commercial model as well.

This session offers a grounded, honest view of what the business case for AI in process manufacturing looks like – what’s working, what was harder than expected, and where the industry still has real work to do.

Annie Lu

CEO & Co-Founder
Laminar

In this session, Omron Senior Product Management will share a practical roadmap for creating a data-first culture that reduces the cost and complexity of deploying advanced process controls. While much attention has been given to algorithm development and deep neural networks, this session focuses on the critical foundation for successful AI adoption: factory operations.

Manufacturers face unprecedented pressures: volatile supply chains, unpredictable consumer demand, and relentless shareholder expectations. Competitive advantages increasingly depend on flexible yet consistent production, and the key to achieving this lies in leveraging AI strategically, not as a future concept, but as a tool built on robust data, magnifying current competitive advantages.

This session will guide attendees through three milestones of advanced process control maturity:

1. Foundational Data – How to capture high-value process insight non-intrusively.
2. Prototype AI – How to create robust designs for edge AI deployment.
3. Scaled AI – How to build up from the edge to cloud-based solutions without ripping and replacing.

By framing these concepts within operational efficiency, the session will illustrate how an intentional data-first strategy can strengthen production systems today while preserving flexibility for advanced process control tomorrow. Attendees will leave with actionable steps to maximize uptime today through non-intrusive data collection and be able to avoid costly challenges when seeking to build a scalable intelligent automation architecture.

Thomas Kuckhoff

Senior Product Manager
Omron Automation Americas

Most manufacturers rely on end-of-line inspection, yet relatively few inspect products at the manufacturing steps where defects actually occur. The challenge is no longer whether AI can detect defects—it can. The real barriers are deploying AI quickly, training reliable models with limited production data, and achieving a return on investment that justifies inspection at every critical process.

This session explores the fundamental differences between end-of-line and in-process inspection, explaining why they require different deployment strategies and technologies. End-of-line inspection prioritizes maximum detection accuracy and can tolerate weeks of engineering and integration. In-process inspection, by contrast, must be deployed in hours, trained with only a handful of sample images, and adapt rapidly to changing production conditions without disrupting throughput.

Attendees will learn how advances in Generative AI, sample-efficient learning, and 2.5D imaging are making scalable in-process AI inspection practical for reflective, high-mix manufacturing environments. Through real-world deployment examples, the session will demonstrate how manufacturers can reduce scrap, rework, and labor costs while improving yield and accelerating time to value. Participants will leave with a practical framework for selecting the right inspection strategy and deploying AI inspection from pilot projects to production-scale implementation.

Keven Wang

CEO
UnitX Inc

As manufacturers continue to increase automation and inspection requirements, machine vision systems are being used in a growing range of production environments where speed, variability, and reliability create significant implementation challenges.

This presentation discusses practical experiences from applying 3D laser triangulation and smart thermal imaging technologies in industrial inspection and process monitoring applications. Examples will include semiconductor inspection, automotive manufacturing, food production, and other high-throughput production environments.

The session will focus on technical and operational considerations that influence the success of vision deployments, including:

• Managing tradeoffs between speed, resolution, and robustness
• Working with reflective, low-contrast, or variable materials
• Synchronization and multi-sensor integration challenges
• Calibration and stability in changing production conditions
• Thermal imaging considerations for process monitoring applications
• Practical limitations encountered during deployment and commissioning

Several application examples will be used to illustrate how different system architectures and sensing approaches were selected based on the specific inspection or measurement requirements.

The presentation will also touch on broader industry trends, including where AI-enabled inspection and advanced sensing are showing practical value today, as well as areas where implementation remains difficult or highly application-dependent.

The goal of the session is to provide attendees with a grounded engineering perspective on deploying advanced vision technologies in production settings, including lessons learned, common challenges, and considerations that can help guide future projects.

Gretchen Alper

Business Director, North America
AT-Sensors

Rashmi Misra

Supervisory Board
Mercedes-Benz Group AG

For robots to operate safely and effectively in real-world environments, two capabilities must come together: perception and trust. Robots need to understand the physical world in 3D—and they need to understand who they are interacting with.

This session introduces a dual-layer architecture for physical AI systems:

• A perception layer (“Vision Cortex”) that enables spatial awareness through depth sensing and multimodal vision
• An identity layer that enables secure, personalized, and accountable interactions with humans

Based on lessons from real deployments, we explore how these layers work together to unlock new capabilities in industrial and service robotics—from safer human-robot collaboration to authenticated task execution and delivery.

Topics include:

• Why perception alone is not enough for real-world autonomy
• Integrating depth sensing with identity-aware workflows
• System design considerations: latency, edge processing, and reliability
• Privacy and governance challenges in vision + identity systems
• Practical deployment patterns and what breaks at scale

This talk offers a forward-looking but grounded perspective on how perception and identity together form the foundation of trusted autonomous systems.

Chris Matthieu

VP, Developer Ecosystem
RealSense, Inc.

Jeff Adolf

Vision/AI Specialist
3M

Ricky Watts

General Manager and Sr. Director, Industrial and Robotics Division
Intel Corp.