Machine-learning running over video is changing the game in manufacturing, making it possible to inspect, monitor, and serve as a digital fingerprint for production traceability.  For decades machine vision & rules-based algorithms have been applied to automate visual inspections using single shot images from industrial cameras.  In recent years companies have started to use machine-learning models with those images to better classify defect types in hopes of mimicking the way humans manually inspect.  Single-shot images are not ideal for all manufacturing inspection/monitoring use cases.  In this talk we will explore what different use cases are better suited for video and why by quantifying the capabilities of video-based computer vision.  We will compare and contrast video-based techniques and discuss where video outperforms still images, but where video might not be the best suited tool.  

Speakers

Adam Bennett Director of Enterprise Sales Matroid, Inc. (United States)

Adam Bennett is Director of Enterprise Computer Vision at Matroid, where he leads deployment, development, and scalability of computer vision across manufacturing, logistics, and hyperscaler operations worldwide. With ~15yrs of experience in deploying and developing new 2D/3D imaging technologies and track & trace technologies in manufacturing he has delivered advanced solutions across many manufacturing sectors such as aerospace, electronics, automotive, life sciences, materials, and more. Over the last 5yrs Adam has been focused on state-of-the-art computer vision for the industrial sector.

Patrick Wang Deep Learning Field Engineering Manager Matroid, Inc. (United States)

Patrick Wang leads the delivery of deep learning and computer vision solutions at Matroid, supporting applications across manufacturing, aerospace, and transportation.

With seven years of experience in applied machine learning spanning academia and industry, he has directed large-scale deployments that improved operational efficiency and secured multimillion-dollar renewals. As head of Matroid’s Deep Learning Field Engineering team, Patrick oversees customer engagements from scoping and system architecture through integration, while contributing to RFPs, technical documentation, and pricing strategy.

His work focuses on translating cutting-edge AI research into scalable, real-world systems. He is first author of a paper at ICLR 2021, serves as a reviewer for NeurIPS and ICLR, and is a co-inventor on multiple U.S. patents in vision-language modeling and video summarization.

At Matroid, Patrick has led projects including PPE compliance and flame detection for workplace safety, refractory-brick and precision-metal defect inspection, cycle-time analytics on production lines, non-intrusive vehicle screening for security, and engine inspection for space-exploration applications.

 

---

Event Details

FORMAT: Oral presentation followed by audience Q&A.
MENU: Coffee, decaf, tea and water will be available nearby.
SETUP: Theater seating.