01Image segmentation and scene reconstruction
Break complex scenes into meaningful regions, and build 3D models of real environments from visual input — for digital twins, spatial analysis and training.
AI serviceCameras and drawings · CV-05
Your cameras and drawings are sensors.
Detection, reading and inspection on live camera feeds and engineering drawings.
FeedsExisting CCTV · RTSP
DrawingsDXF · DWG · scanned PDF
RunsEdge or server
AlertsControl room or phone
01/ Overview
Intelligent image and video analysis for real operations
Our computer vision work powers safety systems, inspection and quality assurance, and document digitalisation. Intelligent image and video analysis turns the cameras and drawings you already have into data sources.
Object detection, motion tracking and image classification — customised for energy, manufacturing and industrial IoT, and integrated for real-time use.
- High-accuracy detection across images, video and scanned documents
- Scales from a single feed to large volumes of video streams
- Customisable data preparation, output formats and algorithms
02/ LiveInteractive · runs in your browser
Scan a drawing. Read a gauge.
Two jobs from the same toolkit. On the drawing, run detection: every instrument tag is boxed, read and indexed — click one to find it. On the camera tab, a gauge is read from the needle angle in each frame, with the alarm band checked against the reading.
- Tags detected, read and indexed — clickable
- Gauge value from needle geometry, every frame
- Illustrative drawing and feed — not a client site
03/ What we offer
Inside this service
CV·01
Defect and corrosion detection
Surface anomalies flagged from inspection imagery.
CV·02
Gauge and flare reading
Analog gauges and flare states read from live video.
CV·03
PPE and safety-zone monitoring
Compliance alerts from existing CCTV.
CV·04
Drawing digitalisation
DXF/DWG drawings converted to smart DEXPI format.
CV·05
Tag detection on P&IDs
Instrument tags found and indexed automatically.
CV·06
Engineering document OCR
Datasheets and scans made searchable.
04/ How it works in practice
What you actually get
CV-05·F1
From cameras you already have
No new hardware needed to start — models run on your existing CCTV and RTSP feeds, at the edge or on a server.
- Works with existing CCTV/RTSP
- Edge or server deployment
- Alerts to control room or phone
CV-05·F2
Drawings become data
Dumb DXF/DWG files become DEXPI-format smart drawings with every tag detected, indexed and searchable.
- DXF/DWG parsed to DEXPI
- Tags detected and indexed
- Searchable, chat-ready drawings
CV-05·F3
Tuned to your site
Models are trained on your footage and drawings, with a review loop that drives false alarms down over time.
- Trained on your footage
- Site-specific object classes
- False-alarm review loop
05/ More capabilities
Vision services we also provide
Drag or scroll the strip.
02Video analytics and quality inspection
Track movement, detect patterns and flag defects on production lines and finished goods — from live or recorded footage.
03Access control and site safety
Identity verification at gates and rooms, and real-time activity detection that helps teams respond proactively in critical situations.
06/ Tools and technologyBuilt with proven tooling
OpenCV
PyTorch
TensorFlow
NumPy
Hugging FaceStack
PythonPyTorchOpenCVYOLOTesseract
07/ ProcessHow we deliver
Step 01Feed and drawing audit
Sources, quality and formats assessed.
Step 02Labelling
Site-specific classes annotated.
Step 03Model training
Detection tuned to your imagery.
Step 04Site pilot
Live trial with alert review.
Step 05Rollout
Scaled across feeds and sites.
08/ MethodSMHcoders
“Your cameras and your drawings are sensors. Vision models turn both into live, structured data.”
SMHcoders · method
Free 45-minute consultation with an engineer: bring one problem, leave with a feasibility read.
09/ Common questions
Do we need new cameras?
Usually not. Models run on your existing CCTV and RTSP feeds; we assess feed quality in the audit and tell you if a viewpoint genuinely needs hardware.
What data do we need to start?
Most engagements start with 6–24 months of historian, lab or document data. We audit coverage in the first week and tell you plainly if it is not enough — before you commit.
How long until we see results?
A proof of concept on your historical data typically lands in 4–8 weeks. Production deployment follows once the KPI target is met and your team signs off.
Where does it run, and who sees our data?
On your infrastructure — plant-side servers, your Azure or AWS tenancy, or an air-gapped network. Data never leaves your boundary, and nothing is used to train third-party models.
Who owns the result?
You do. Handover includes source code, documentation, training sessions and a retraining plan.
10/ Related services
Proof of concept · 4–8 weeks · your data