First Heterogeneous Embedded Vision Systems Combining Jetson and FPGA

Jan 29, 2023 3 min read

A hand holding the compact FantoVision unit, showcasing Gidel’s first heterogeneous Jetson FPGA embedded vision systems which combine NVIDIA Jetson Orin NX with FPGA acceleration.

Redefining Edge Imaging with Breakthrough Heterogeneous Computing

Big news for vision, imaging, and AI applications. Gidel introduces the first heterogeneous computer that combines NVIDIA Jetson Orin NX with an FPGA-based acquisition and acceleration engine, creating an exceptionally powerful embedded vision system for real-time, high-bandwidth imaging at the edge.

The new configuration brings the NVIDIA Jetson Orin NX 16GB module into the FantoVision20 and FantoVision40 product line. This combination provides up to 157 sparse TOPS, delivering a five-fold performance increase over the previous Jetson Xavier NX generation. As a result, developers gain industry-leading AI throughput, deterministic latency, and real-time image processing capabilities. When paired with Gidel’s FPGA-based frame grabbing, pre-processing, and compression, the platform sustains more than 1 gigapixel per second of imaging throughput.

FantoVision: A Complete Embedded Vision System for High-Speed Imaging

The FantoVision series provides a ready-to-use environment for advanced imaging and AI applications. Each system integrates:

NVIDIA Jetson Orin NX 16GB
Gidel’s FPGA image acquisition and acceleration engine
High-bandwidth camera connectivity
Real-time compression and enhancement options
Compact, industrial-grade mechanics
SWaP-optimized design: 134 × 90 × 60 mm, 750 g

Therefore, developers can deploy real-time and high-resolution vision systems with shorter integration cycles and improved reliability.

Multi-Interface Flexibility for Demanding Vision Applications

The FantoVision20 integrates Camera Link and GigE Vision within the same compact edge computer. Consequently, system designers gain flexibility when selecting the right camera interface for their industrial vision pipeline.

The FantoVision40 model supports high-bandwidth imaging through a combination of 10GigE Vision and CoaXPress 12 connectivity. This allows developers to run high-speed or multi-camera setups with consistent, deterministic performance.

Together, the FantoVision20 and FantoVision40 platforms deliver versatile, multi-interface embedded vision systems that address a wide range of machine vision requirements

Jetson Orin NX + FPGA Acceleration Enables Real-Time AI Vision

The NVIDIA Jetson Orin NX module offers a next-generation CPU, GPU and FPGA architecture with 157 TOPS of AI performance. It runs deep-learning models at the edge with outstanding efficiency.

In addition, the integrated FPGA performs:

High-speed frame grabbing and ROI extraction
Real-time image enhancements such as HDR
Compression (JPEG, lossless, H.264/H.265, Quality+)

This heterogeneous architecture transforms FantoVision into a high-performance embedded vision system capable of sustaining over 1 gigapixel per second through the NVIDIA JetPack SDK.

A Turn-Key Platform for Embedded AI Vision Systems

“The combination of Gidel’s FPGA and image-processing expertise with NVIDIA accelerated computing breaks another barrier between AI and machine vision,” said Reuven Weintraub, Founder and CTO of Gidel.
This heterogeneous architecture, which brings FPGA and GPU processing together in one embedded vision system, gives engineers a turn-key platform for running AI workloads alongside real-time, high-speed imaging.

Gidel will showcase the Jetson Orin NX-enabled FantoVision systems at Embedded World, held in Nuremberg, Germany, from March 14–16, 2023.

Written by

Gidon Weintraub

Gidon Weintraub is a leading expert in machine vision and high-speed imaging systems, serving as VP of Sales and Marketing at Gidel. Having built his entire career at Gidel, he combines deep technical understanding with strategic market insight. Gidon has been directly involved in the detailed architecture, challenges, and bottlenecks of dozens of advanced imaging projects across sorting and inspection machines, defense and military systems, autonomous platforms, high-speed AI vision workflows, electron microscopy, and industrial automation. His unique ability to bridge technical depth with real-world application needs makes him a recognized figure in the global imaging and vision community.