AI Frame Grabber
Edge AI hardware built for fleet deployment.
4 PoE+ GigE Vision inputs with on-board dual NPUs for real-time AI inference. No cloud. No camera replacement. No external GPU. On-card AI inference. Lower latency and lower system cost.
Hardware
Capture and Infer on One Card
Image acquisition and AI inference happen on the same PCIe card — no external server, no added latency from round-tripping frames off-board.
Capture Card
+
CPU
+
External GPU
Consolidated Into
All-in-One
Apexera Card
CaptureDual NPUSecure OTA
1 CardReplaces 3 Devices
7.59 msP99 Latency
16.8 WBoard Power
Traditional systems move every frame to an external GPU.
Apexera Card processes them on the card.
Multi-Camera Processing
Capture and process multiple industrial camera streams simultaneously on a single card.
On-Card AI Inference
Run AI models directly on the frame grabber, reducing data transfers and eliminating the need for a separate GPU.
Fleet-Ready Security & OTA
Secure boot, full-disk encryption, and A/B OTA updates—ready to integrate with your device management platform or operate standalone out of the box.
Where It Fits
Where 4PE Fits in Your Operations
Run multi-camera AI inference directly on the edge across security and industrial environments.
AI Inspection for Smarter Manufacturing
Apexera Card enables real-time visual inspection for high-speed production lines, helping identify defects early with a streamlined edge AI architecture.
AI-Powered Security Monitoring
Apexera Card delivers real-time AI inference for multi-camera surveillance, helping reduce response time and deployment complexity in security and loss prevention systems.
Why Apexera Card
Edge AI Hardware for Fleet Deployment
Apexera Card 4PE helps reduce both upfront hardware investment and the long-term cost of deploying and operating edge AI systems.
Lower Initial Investment
Deploy AI using fewer components and lower hardware investment.
No External GPU
Existing Camera Support
Single PCIe Card
Simplified System Design
Improved Operational Efficiency
Reduce maintenance effort and simplify long-term system management.
Secure OTA Updates
Centralized Management
Easier Troubleshooting
Scalable Deployment
Performance
AI compute on the card
End-to-end measurements on production 4PE hardware. No marketing TOPS.
ResNet-50 v1.5 — Industry CNN Baseline
7.59 ms
P99 Latency
971 SPS
Offline Throughput
~8 passes / frame
@ 4 CAMERAS × 30 FPS
Foundation Models — FP16, no quantization
Model
Latency
Throughput
DINOv3-ViT-S/16 (224)
Latency22 ms
Throughput290 fps
DINOv2-ViT-S/14 (224)
Latency27 ms
Throughput268 fps
DINOv3-ViT-B/16 (224)
Latency60 ms
Throughput114 fps
Vision-language and edge-RAG pipelines on-card. No cloud round-trip.
Internal benchmark on production 4PE hardware. LoadGen v5.1, MLPerf Inference Edge methodology. CNN: INT8 w8a8. Foundation models: FP16. End-to-end measurements include PCIe, DMA, and host-side overhead. Not an MLCommons submission.
Software Ecosystem
Complete AI Vision Solution
Apexera Vision extends Apexera Card 4PE with integrated AI training, deployment, and application management.
Few-Shot AI Training
Train deep learning models with fewer labeled samples, making new vision applications faster to develop and adapt.
Automated Vision Workflows
Configure complete vision workflows from image input and AI analysis to application logic and results.
Seamless Edge Deployment
Deploy models and applications directly from Apexera Vision to Apexera Card 4PE for on-card inference.
Unified Hardware & Software
Configure, deploy, and manage AI applications and 4PE devices through one integrated environment.
Reliability at Scale
Built to Scale
Built for repeatable deployment, evolving AI workloads, and reliable long-term operation.
Standardized Deployment
> Standardized hardware platform
> Consistent deployment across projects
> Repeatable system architecture
Flexible AI Applications
> Support multiple AI applications
> Flexible model deployment
> Adapt to changing vision requirements
Scalable Performance
> Multi-camera AI processing
> Parallel AI workloads
> High-throughput edge inference
Industrial-Grade Reliability
> Designed for continuous operation
> Secure boot and full-disk encryption
> A/B updates for reliable system maintenance
Technical Specifications
The Card, Spec by Spec
Real-time, multi-camera AI vision inspection that integrates seamlessly into your system, improving detection accuracy and quality-control efficiency across industrial automation applications.
Compute & Storage
Product Model
4PE
CPU
Rockchip RK3588
AI Accelerator
RK1820
Memory
8 GB
Storage
64 GB eMMC
Bus Interface
PCIe x4
Connectivity & Cameras
PoE
4× IEEE 802.3at PoE+
(25.5 W / port · 50 W total)
(25.5 W / port · 50 W total)
Ethernet
4× 2.5 GbE
I/O
1× CAN · 1× GbE · 1× USB 2.0
· 1× USB‑C · 1× DisplayPort
· 1× USB‑C · 1× DisplayPort
Serial
1× RS‑232 · 1× Isolated RS‑485
Supported Cameras
GigE Vision
Max Cameras
4
Software & OS
Software Platform
Apexera Vision
(automation with deep‑learning support)
(automation with deep‑learning support)
OS
Windows・Linux
Power, Mechanical & Environmental
Power
12 V DC
Dimensions
227.7 × 111 mm (W × H)
Operating Temp.
0 °C ~ +60 °C
Storage Temp.
0 °C ~ +60 °C
Humidity
10–90%, non‑condensing
FAQs
Frequently Asked Questions
Don’t see your question answered here? Contact us today.
What is an Apexera Card?
An Apexera Card is a PCIe image acquisition card with integrated AI inference. It combines image capture and AI processing on a single hardware platform for real-time industrial vision and edge AI applications.
How is an Apexera Card different from a traditional frame grabber?
Unlike a traditional frame grabber that only captures images, an Apexera Card performs both image acquisition and AI inference on the same PCIe card, reducing system complexity and enabling real-time AI processing.
Can I use my existing industrial cameras with Apexera Card 4PE?
Yes. Apexera Card 4PE is designed to work with existing industrial cameras, allowing AI to be added without replacing your current imaging infrastructure.
Which AI applications can Apexera Card 4PE support?
Apexera Card 4PE supports a wide range of machine vision applications, including defect detection, OCR, barcode reading, object detection, intrusion detection, intelligent surveillance, and visual inspection.
Which operating systems and AI frameworks are supported?
Apexera Card 4PE supports Windows and Linux, along with industry-standard AI deployment workflows. Refer to the technical specifications for the latest framework compatibility.
Get Started
Ready to Deploy AI at the Edge?
Whether you’re building a new vision system or upgrading an existing one, we’re here to help you find the right configuration.