FantoVision20 Embedded Computer Vision System
The ultra-compact FantoVision20 embedded vision computer delivers real-time image acquisition, compression, and processing at up to 20 Gb/s. Designed for embedded computer vision workloads, it enables high-speed, high-resolution imaging in systems with strict SWaP limits. As a result, it serves demanding industrial, mobile, and outdoor applications.
For more details regarding the FantoVision20 edge AI system, visit: FantoVision20
Ultra-compact embedded computer vision system for 20 Gb/s imaging
Gidel, a technology leader in high-performance FPGA systems, introduces a small and robust platform tailored for embedded computer vision and high-throughput processing. FantoVision 20 allows image processing, compression, and recording of video streams up to 20 Gb/s in real time. It delivers this performance in an extremely compact form factor, enabling integration where size and power budgets are limited.
High-Speed Image Acquisition and Real-Time Processing
With 2 × 10GigE and Camera Link 80-bit (DECA) inputs, FantoVision20 provides high-bandwidth camera interfaces. This allows the system to capture and process high-resolution images at fast frame rates in real time.
FantoVision20 can acquire directly from high-bandwidth GigE Vision cameras and Camera Link cameras, supporting flexible embedded vision architectures for high-speed and multi-camera imaging.
The architecture combines NVIDIA Jetson computing for AI and application processing with an Altera Arria 10 FPGA for deterministic image acquisition, preprocessing, and compression. Developers can use C/C++, CUDA, and the NVIDIA JetPack software environment for AI and application development. Meanwhile, the FPGA enables advanced preprocessing and FPGA Image Compression IPs for real-time data reduction. With multiple FantoVision20 units and Gidel’s open-FPGA InfiniVision flow, integrators can capture and synchronize more than 1,000 camera streams in scalable systems.
Thanks to its 134 × 90 × 60 mm³ form factor, FantoVision20 fits easily into tight installations. Its rugged housing withstands harsh industrial environments, and both passive and active cooling configurations are available.
Embedded Computer Vision Applications
With its strong processing capabilities and compact design, FantoVision 20 enables new embedded computer vision deployments. Typical applications include high-speed inspection, ITS, broadcast, medical imaging, agriculture, and aerial mapping.
High-bandwidth interfaces and real-time processing also enable sorting and inspection tasks at high resolution. A 10GigE switch can connect multiple GigE cameras for 360° inspection or 3D scanning. NVIDIA AI tools support machine learning and inference pipelines. Due to its compression support and size, it can also serve as a portable offline recording system.
For surveillance or traffic monitoring, FantoVision 20 enables compression, local recording, and streaming. Its compact housing installs easily on poles or gantries, and long cable interfaces allow flexible placement relative to cameras.
The lightweight, low-power design also supports vehicle-based systems, such as 360° mapping, agricultural vision tools, weed detection, targeted fertilization, and autonomous harvesting.
With high-bandwidth camera interfaces, NVIDIA Jetson AI processing, and FPGA-based acquisition and preprocessing, FantoVision20 supports demanding embedded computer vision applications while remaining compact and SWaP-efficient.
