Newsroom > Blog
Aug 6, 2026
Embedded Vision Summit 2026 held on May 11–13, 2026 in Santa Clara, CA, connected 1,400+ product and application developers, business leaders, investors, and customers focused on physical AI at the edge. Featured sessions included an Efinix presentation from our VP of Marketing, Mark Oliver, titled “Why Your Next AI Accelerator Should Be an FPGA” along with other tutorials and case studies.
Super Resolution and Segmentation generated the most engagement. Visitors were interested in how these demos could improve image quality, support real-time detection, and run efficiently at the edge. Super Resolution was relevant for industrial inspection and image magnification, while edge AI use cases such as security cameras benefit from clearer facial details and license plate visibility.
Visitors also showed interest in video connectivity, low power, small form factor, and practical use cases for camera-based systems.
Most questions focused on power consumption, frame rate, resource usage, and how the demos could be used in real applications. Visitors responded well to the visible low-power setup, especially since the demos did not require heat sinks.
Demo #1: Super Resolution
The Super Resolution demo showed how lower-resolution images can be upscaled to improve visual clarity. This is useful when you need to magnify a small pixel window without losing resolution. Common use cases include industrial inspection, image magnification, and security cameras where improved clarity can help identify faces or license plates.
The demo used QuickSRNet to scale image input from 188x144 to 720x576 and reached 83 FPS on the Ti375C529 platform. Super Resolution helps improve image quality for edge vision systems while keeping the design efficient and compact.
Demo #2: TinyML / Multicore
The TinyML / Multicore demo showed how lightweight machine learning workloads can run efficiently on Efinix FPGA devices. This is useful for edge applications that need fast local processing in a small, low-power design. TinyML / Multicore shows how Efinix devices can support efficient edge processing for compact embedded systems.
Demo #3: Segmentation
The Segmentation demo showed real-time person detection and masking. Instead of only drawing a box around a person, segmentation identifies the person’s shape within the image, making the output more useful for vision-based applications. While this demo focused on people detection, models with more parameters could classify additional objects such as bikes, cars, and other scene elements.
The demo used Yolov5n-Seg for instance segmentation and ran at 40 FPS on the Ti375C529 development kit. Potential use cases include crowd monitoring and event security, automotive traffic management on roads and intersections, and industrial inspection such as PCB component placement and identification.
Demo #4: SLVS + SDI Connectivity
The SLVS + SDI connectivity demo showed how Efinix FPGAs can support reliable video transmission and signal conversion. This type of demo is useful for camera, display, and video pipeline applications that need high-quality data movement.
Video connectivity shows how Efinix FPGAs can support professional video, embedded cameras, and display-based systems.
The main takeaway was that Efinix FPGAs can support practical embedded vision and video applications with high performance and low power. The demos showed that real-time processing can happen directly at the edge without needing large, power-heavy systems.
Key points included low power, small-form-factor packages, integrated memory SiP packages, and fast performing fabric for running AI at the edge.
Mark Oliver, VP of Marketing at Efinix, presented “Why Your Next AI Accelerator Should Be an FPGA” at the May 2026 Embedded Vision Summit. The presentation focused on why FPGAs are a strong fit for edge AI systems, especially when cost, latency, complex I/O, and tight power budgets matter.
Visitors responded well to the low-power and small-footprint message. They liked seeing working demos that connected to real-world applications, especially where size, power, performance, and reliability matter.
The demos are useful for applications such as smart cameras, security systems, traffic monitoring, robotics, industrial inspection, medical imaging, automotive systems, and professional video equipment.
Embedded Vision Summit was a good fit because the event focuses on embedded vision, edge AI, and real-time visual processing. These areas align well with Efinix FPGA solutions, especially for customers looking for low-power, compact, and flexible platforms.
Efinix technology was also represented through partner activity at the event, including Winbond’s external HYPERRAM memory solution shown alongside an Efinix FPGA board for compact edge and IoT applications.
FPGAs are useful for edge vision because they can support real-time processing, flexible I/O, and efficient performance in a small form factor. This makes them a strong option for applications that need fast image or video processing close to the sensor.
The event gave Efinix a strong opportunity to show practical demos, connect with customers, and highlight how its FPGA technology supports real-world embedded vision and edge AI applications. It also helped show how Efinix solutions can be used across camera, video, industrial, and smart vision systems.