Integrate vision AI in latency-critical operations

Robovision Edge is the missing link to deploy vision intelligence to production. This high-security edge device serves predictions within milliseconds to your on-premise solutions powered by models created, maintained and continuously improved on the central Robovision Platform.

Talk to an expert

Bridge the last mile of AI delivery

Deploying and maintaining AI models in production is where most AI projects go wrong. Functioning as the local brain for on-premise needs, Robovision Edge makes AI really happen by capturing, processing and uploading images. Then it seamlessly integrates with the Robovision platform for model deployment and adaptation, all in a streamlined and scalable workflow with our edge device.

No complication from the get-go


1. Install the Robovision Edge device and cameras

Mount the Robovision Edge device next to the production line. Connect the power supply and up to three cameras under a consistent angle and lighting, depending on your production line configuration.


2. Run the application

In the device interface, set up the camera and choose which AI model to deploy. Start the monitoring. Your AI-powered computer vision application is now up and running.


3. Continuously improve your models

Model retraining needed? Set up the cameras to record data in the device interface. Upload these training images to the Robovision AI Platform and integrate the retrained model in the device afterwards.

Managing an end-to-end AI life cycle

Robovision Edge connects with the Robovision Platform that covers the full AI lifecycle for latency-critical operations

From gathering data to deploying models, our Robovision Platform and Robovision Edge devices cover every step of the AI lifecycle. Quickly output new models for new products and maintain AI solutions yourself.

Closing the AI Life Cycle Loop

Together with our platform, you can now create, update, and deploy AI in a continuous cycle: Perform data preparation and model training on the Robovision AI Platform, and data sourcing and model deployment on the Robovision Edge device.

Low Latency with High-performing GPUs

Run your applications at low latency with the NVIDIA® Jetson™ Xavier AGX module. Its 32 TOPS of AI performance enables the edge device to simultaneously run up to three cameras to capture, process, store, and upload images at a combined 40 (FPS) frame rate.


Harvest the value of your retrained AI model by integrating it with your PLC infrastructure through the built-in OPC-UA interface. Then selecting a model in the Robovision Edge interface is all it takes to deploy AI into your factory processes.

Industrial-grade Hardware

The device’s robust and fanless design enables it to operate anywhere between ‑10 and 60 °C, in dusty or humid environments. Robovision’s compact and ruggedized edge device slots right into most factory configurations.


A different hardware set-up for each production line—that’s a thing of the past. Streamline the integration of AI across lines and factories with one, standardised edge device model.


From the ground up, Robovision Edge was designed for scaling. Setting up a Robovision Edge device for each new line is all you need to do to expand your AI ambitions.

Out-of-the-box algorithms for standard AI processes

Robovision platform supports standard algorithms such as image classification

Unlike traditional machine vision, modern vision AI is capable of replicating the human visual system. It enables high-accuracy quality control and the automation of complex manufacturing tasks. The Robovision Platform allows production operators to take control of any edge solutions they collaborate with in an ever-changing dynamic environment.

Robovision platform supports standard algorithms such as object detection

Recognising different objects is a fairly easy task for any human. At last, we’re making it just as easy for machines. With the onset of deep learning, you can now easily—and accurately—identify and locate objects using photos or videos. From microscopic flaws to jaywalkers, our AI can be trained to see exactly what you want them to see.

Robovision platform supports standard algorithms such as image segmentation
Image segmentation is the task of clustering together parts of an image that belong to the same object class. It accelerates the development of real-world applications across industries, including gas cylinder detection or detection of Covid-19 disease from X-ray images.

Hassle-free integration

Speed and efficiency is key in production, real-world demands that Robovision Edge is designed for. The device’s powerful GPU makes fast, real-time processing on the conveyor possible. And its integration with the Robovision AI Platform gives it the flexibility to efficiently adapt AI to changing data or business environments.

Boost your production process with faster AI delivery

Use the Robovision Edge device for all kinds of computer vision applications, whether you need to detect gas bottles or spot fabric defects.


In-line quality control

AI for defect monitoring is here to stay. AI-powered visual inspection applications for quality control have already surpassed human capabilities. They ceaselessly deliver consistent output whereas human workers get tired and lose focus. This reduces costs, frees up manpower and ensures highly consistent output quality.

Food sorting machinery

Agriculture is becoming a high-tech sector. Workers who used to manually sift through produce are now at the helm of intelligent food sorting machines. Automation will be key towards solving challenges such as food waste and experienced labour shortages.

Technical specifications

With a high-performing GPU, a flexible port configuration and efficient power consumption, Robovision Edge is industrial-grade hardware designed to operate in harsh industrial environments.

Dimensions (W x D x H) 192 x 230 x 87 mm (7.50 x 9.05 x 3.42 in)
Weight 4.5 kg (9.9 lbs)
Installation methods Desktop and wall-mounted fitting
Capturing frame rate 1 camera: 40 fps
2 cameras: 20 fps per camera
3 cameras: 13.3 fps per camera
GigE & USB 3.0 camera compatibility Basler Pylon SDK-supported RGB and mono area scan
Allied Vision Vimba SDK-supported  RGB and mono area scan
IDS uEye(+) SDK-supported RGB and mono area scan
Inference throughput Object Detection (Yolov5): 20 fps at a resolution of 640x640x3
Classification (EfficentNet): 40 fps at a resolution of 224x224x3
Internal storage capability 1 TB
Jetson AGX Xavier GPU NVIDIA Volta architecture with 512 NVIDIA CUDA cores and 64 Tensor cores. 11 TFLOPS (FP16). 22 TOPS (INT8)
Industrial communication OPC-UA
2 x USB 2.0 and 2 x USB 3.0
16 bit (8-In 8-Out) DI/DO
COM: 2 x RS-232/RS-422/RS-485
2 x ethernet port (RJ45) at 10/100/1000 Mbps

Move AI to the real world

Talk to Us Today