Visual inspection at line speed, trained by your own teams.
Detect defects and safety risks directly on the production line, in under 50 ms per image. Your quality experts label the images, Newboot handles training, deployment and monitoring at the edge.
Human visual inspection cannot keep up with machine speed.
The line runs around the clock at the same pace. Attention does not. Most vision projects then stall when it is time to go from one pilot to every line.
Quality varies from shift to shift
Scrap rates fluctuate even when settings are stable, because manual inspection depends on fatigue and experience.
Safety is checked after the fact
Missing PPE or a restricted zone crossed shows up in the next day’s report, not when it happens.
Every new site starts from zero
A vision project that works on one line rarely moves to the next one without a new integration project.
From the first image to a model on the line, in one workflow.
Operations and quality teams build their own use cases. No data science team required. Select a step to see it in the product.
Figures from our gear segmentation demo scenario, not customer data.
The model runs next to the camera, not in the cloud.
Inference happens on an industrial edge device installed on the line. Results go straight to the systems your operators already use.
One platform for quality, safety and production.
Find defects at machine pace
Surface, geometry and assembly defects detected on every part, with automatic traceability of each inspection.
Prevent incidents instead of reporting them
Missing PPE (helmet, glasses, gloves) and restricted-zone intrusions detected in real time, connected to your shop-floor alarms.
Check what the eye checks today
Read labels, verify assembly steps and count parts directly from the camera stream.
Proven on demanding materials
Deployed on glass fiber, technical textiles and building materials lines, where defects are small, fast and varied.
Automated defect detection on an innovative textile line, rolled out across sites.
Adfors, a Saint-Gobain company, needed to detect several types of defects in real time so operators could step in immediately, without stopping the line or launching an IT project.
How it was deployed
- The Newboot edge computer acquires the video streams of 4 industrial cameras in real time.
- Inference runs on the edge, in parallel for each camera, triggered by an OPC-UA tag from the line.
- Results and a low-res image are published on the edge device’s OPC-UA server.
- The Ignition MES shows each detection to operators in real time.
First deployed in Neustadt (Germany), replicated in Buggenhout (Belgium), then in Mexico. Rollout to 36 more sites is under way.
Built to replicate, not to restart.
Reuse
Models, annotations and best practices carry over to the next line.
Standardize
The same methods, connectors and deployments on every site.
Monitor
Follow models and performance across all your sites from one interface.
Extend
Add use cases, lines and sites on the same foundation.
Questions we hear from plant teams.
Do we need a data science team?
No. Quality and operations teams label images and run training from a guided workflow. Newboot supports the first use case and the integration with your systems.
Do images leave the plant?
No. Inference runs on an edge device installed on site. Images are processed where they are captured and detection keeps running without a cloud connection.
Does installation stop production?
No. Cameras and the edge device are installed on a running line, without changing the production process.
Which systems can receive the results?
Results are published over OPC-UA, MQTT or API to your MES, SCADA, ERP, historian or alarm system. At Adfors, detections are displayed in the Ignition MES.
How fast is detection?
Under 50 ms per image on the edge device, fast enough to act while the part is still on the line.
Can we deploy the same model on other sites?
Yes. A model trained on one line is deployed remotely to other lines and sites from a central interface, as Adfors did from Germany to Belgium and Mexico.
Let’s find the right starting point together.
We start with a workshop to assess your use case and define a high-value pilot.
- ?Which quality, production or safety issue costs you the most today?
- ?Can it be detected visually?
- ?What measurable gain would justify a larger rollout?