Detect
Recognise objects, codes, products and gestures. Extract required information from a frame.
Computer Vision
See more.
Act more precisely.
We develop computer vision for business and industry, turning images and video into information your systems can use.
Frame → Information → Action
Vision for your processFrom research to integration
Images become informationDetect.
Detect.
Understand. Track.
Select methods and neural network architecture for your task. Capture conditions, object types and the intended use of results matter.
Understand
Segment images and analyse object characteristics, from defects to terrain.
Track
Count objects and track movement. Turn zone crossings or changes into events.
Three computer vision tasksAn object in the frame.
An object in the frame.
Meaning in the system.
Select a task to see how the processing output changes for the same image.
Recognise and classifyProcessing diagram
Illustrative frame
ObjectWhat is in the frame
CategoryWhich class it belongs to
Connect the result to an action: accounting, verification or notification.
Identify the required areaProcessing diagram
Illustrative frame
SegmentationObject boundaries in the image
AnalysisDefect or distinctive area
Work with specific regions: products, defects or objects in satellite imagery.
Track movementProcessing diagram
Illustrative frame
TrajectoryWhere the object was and where it is moving
EventZone crossing or change
Track movement, count objects and record agreed events.
Industries and workflowsYour process.
Your process.
A vision workflow of its own.
We work across industries. Criteria and acceptable errors are defined before development.
Retail
Shelf stock, queues, product placement and visitor paths. Analysis helps plan store operations.
Banking and security
Video surveillance and access monitoring for restricted areas. Events and requirements are defined for the project.
Manufacturing
Defect detection, production process monitoring and safety rule compliance.
Healthcare
Research analysis of medical images with domain specialists. Solution requirements are developed separately.
Agriculture
Identify damaged products and remotely observe plants and animals.
Your industry
Unusual objects and conditions. We review the task, test samples and propose an approach.
The basis for a pilotTest on what
Test on what
your camera sees.
Quality on an ideal image does not establish quality in your process. Validation includes normal and challenging conditions.
Materials from
your process.
Study real images and video, considering light, angle, movement, occlusion and typical defects.
- Object samples and difficult cases
- Capture parameters and video sources
- Labels and classification rules
Criteria
of your output.
Agree which errors matter most and what action the system should take after processing.
- Quality on an agreed dataset
- Speed and required workload
- Events and integration method
Team experienceChallenging images.
Challenging images.
Real projects.
Computer vision and image preparation experience from the MST portfolio.
Industrial recognition
QR codes in streaming video
Developing and improving production line marking recognition.
Read the project
Data preparation
Earth remote sensing
Creating datasets to analyse the Earth's surface.
Read the project
Production markingWorking with codes?
Working with codes?
Meet MST ScanVision.
Our software recognises QR and Data Matrix codes in video. Other tasks use separately designed solutions.
Technology for the taskSelect the approach.
Select the approach.
Validate the result.
We use tools for models, images, data analysis and visualisation.
01 /TensorFlowMachine Learning
02 /KerasModel development
03 /scikit-learnAlgorithms and metrics
04 /OpenCVImages and video
05 /pandasTabular data
06 /NumPyNumerical computing
07 /SciPyScientific computing
08 /MatplotlibVisualisation
How we workA clear process.
A clear process.
A verifiable result.
Start with research or commission development and deployment.
-
01
Introduction
Discuss the task and samples. Sign an NDA if required.
-
02
Analysis and planning
Review capture conditions, data and requirements. Agree validation and solution scope.
-
03
Agreement
Record terms, timing and responsibilities in the contract.
-
04
Development and launch
Build the solution, test workflows and integrate outputs.
Questions answeredDiscuss the essentials.
Discuss the essentials.
Before launch.
Can we use existing cameras?+
Compatibility and image quality are assessed using equipment specifications and samples. Lighting, angles or the source itself may need changes.
What other tasks can we discuss?+
Product and price recognition in receipts, gestures, satellite imagery objects, passenger counting and restricted-area events. Specialised facial expression analysis is discussed as a separate research task with verifiable criteria.
Can accuracy be stated in advance?+
Quality depends on data and conditions. We agree a dataset and criteria, then measure pilot results. We do not promise one accuracy level for all objects.
How will results reach our system?+
We design integration interfaces and output formats for your process. These may be recognised features, objects or events for further processing.
Start with your challengeShow us the task.
Show us the task.
We will propose an approach.
Tell us what to detect, recognise or track. Start with your samples and conditions.
- Objects and expected actions
- Capture source and conditions
- Quality and speed requirements
Solution scope, timelines and terms are agreed for your project.
Loading the contact form…