Gidel at Embedded World 2024

Embedded World 2024: Gidel Demonstrated Embedded Vision and Edge AI

At Embedded World 2024, Gidel demonstrated FPGA-based embedded vision technologies for high-resolution acquisition, real-time image processing, and Edge AI.

The demonstrations showed how Gidel combines high-bandwidth camera acquisition, deterministic FPGA processing, and NVIDIA Jetson computing in compact embedded systems for demanding imaging and vision applications.

FantoVision Embedded Vision with FPGA Acquisition and NVIDIA Jetson

Gidel presented its FantoVision Edge AI Systems, which combine NVIDIA Jetson computing with FPGA-based image acquisition and real-time processing.

FantoVision supports high-bandwidth imaging interfaces including GigE Vision, CoaXPress, and Camera Link. The FPGA handles deterministic acquisition and pixel-intensive processing before transferring image data to the Jetson for AI and application processing.

This architecture reduces acquisition and image-processing overhead on the embedded processor, leaving more CPU/GPU resources available for inference, detection, tracking, classification, and application logic.

High-Bandwidth Camera Acquisition for Embedded Vision

Gidel’s embedded vision architecture supports multiple machine vision interfaces for different bandwidth and system requirements:

• GigE Vision frame grabbers for flexible high-bandwidth Ethernet-based acquisition
• CoaXPress frame grabbers for very high-bandwidth acquisition, including CXP-12
• Camera Link frame grabbers for deterministic industrial camera acquisition

Engineers comparing camera interfaces can also review Gidel’s guides to CoaXPress cameras, GigE Vision cameras, and Camera Link cameras. The guides cover bandwidth, connectivity, synchronization, cabling, and acquisition considerations.

Real-Time FPGA Processing for Edge AI

Gidel demonstrated how FPGA processing can offload image-processing workloads from NVIDIA Jetson processors.

The FPGA can perform functions such as image correction, enhancement, HDR, compression, protocol processing, and data preparation directly within the acquisition pipeline.

This enables deterministic real-time processing with low added latency while preserving Jetson CPU/GPU resources for AI inference and application workloads.

Developers who need custom FPGA processing can use Gidel’s ProcVision Suite to develop and integrate their own FPGA imaging algorithms.

Real-Time HDR and Quality+ Compression

Gidel also demonstrated FPGA-based image enhancement and data reduction technologies at Embedded World 2024.

Real-Time HDR IP improves image visibility in scenes with challenging lighting and high dynamic range while operating directly within the FPGA processing pipeline.

Quality+ Compression reduces image data volume in real time while maintaining high image quality, helping reduce downstream bandwidth and storage requirements.

Embedded Vision Applications

• Embedded AI and machine vision
• Industrial inspection and automation
• Multi-camera imaging systems
• High-bandwidth image acquisition
• Real-time FPGA image enhancement
• Edge AI detection, tracking, and classification
• Compact and SWaP-constrained imaging systems

Explore Gidel Embedded Vision Platforms

• FantoVision Edge AI Systems
• PCIe Frame Grabbers
• Mini FPGA Modules

Embedded World 2024 Event Details

When: April 9 to 11, 2024
Where: Nuremberg, Germany
Booth: 2-559

Official event website: Embedded World