Real-time industrial computer vision that detects defects, blockages, and process deviations the moment they form — triggering automated plant responses in under a second.
In-Sight Vision is a purpose-built industrial computer vision system. It trains deep learning models on your specific environment — your materials, your machines, your failure modes — and deploys them at the edge for sub-second, always-on detection.
Unlike off-the-shelf vision systems, In-Sight Vision is custom-trained from the ground up on your site's data. The models learn what normal looks like in your operation — and alert the moment something deviates.
And it integrates directly with your plant control systems — so detection isn't just a dashboard notification. It triggers the right automated response, immediately.
Live camera feeds are connected to the edge inference system — covering every critical point in the process where visual inspection is needed.
The trained deep learning model analyses every frame in real time — identifying the specific objects, defects, or conditions it was trained to find.
When a detection event occurs, the system triggers configurable alarms — dashboard alerts, SMS, email, and direct PLC integration for automated plant response.
New detections are captured and fed back into the model training pipeline — continuously improving accuracy as your process evolves.
In-Sight Vision handles the full computer vision lifecycle — from custom model training to edge deployment and continuous improvement.
Every deployment is trained from scratch on your site's images. We capture, annotate, and train a model that understands your specific materials, equipment, and failure signatures.
Models run at the edge — directly at the camera and point of inspection. No round-trips to the cloud. No latency. Detection in under 1 second, even in low-connectivity environments.
Detection events trigger direct signals to your PLC — enabling automated plant responses like conveyor stops, alarms, or divert gates. No operator intervention required.
We operate end-to-end on the training pipeline. At Richards Bay Minerals, we collected and annotated 40,000+ training images — giving the model the variety it needed for high real-world accuracy.
The system never takes a break. No shift handovers, no fatigue, no missed detections. Every frame, every hour, inspected with the same precision — day and night.
All detection events are logged, timestamped, and visualised in a live dashboard. Trend analysis, frequency reports, and historical review — giving you visibility into how your process is performing.
Richards Bay Minerals faced up to 4-day production shutdowns from undetected calcrete blockages forming in their pond pump operations. The problem was invisible until it was too late.
Excite-Data deployed In-Sight Vision on the pond — training a custom model on 40,000+ annotated images and integrating directly with the plant's PLC infrastructure. The system now detects calcrete formation in under 1 second, 24/7, triggering automated responses before blockages form.
Read the Full Case StudyAny industrial process where visual inspection currently relies on a human eye is a candidate for In-Sight Vision.
Detect pipe blockages, pond obstructions, and conveyor pile-ups before they halt production. Proven at Richards Bay Minerals.
Identify surface defects, dimensional deviations, and assembly errors at line speed — with consistent accuracy that no human inspector can match across an 8-hour shift.
Automate cut-point identification on rolling mills and casting lines — recovering material and eliminating operator-dependent variation. Delivered R11M savings per year.
Automated label verification, barcode reading, and package integrity checks at conveyor speed — eliminating manual QC bottlenecks in dispatch and receiving.
Real-time detection of PPE compliance violations — hard hat, vest, and footwear — in restricted zones, triggering alerts before incidents occur.
Smoke, fire, overflow, liquid spills — visual anomalies that indicate process deviation. Detected automatically, before the damage compounds.
Book a discovery call. Tell us about your process and what you're trying to detect — we'll tell you honestly whether In-Sight Vision is the right fit and what a deployment would look like.