Custom computer-vision models
Evaluate, adapt or develop models for defined image classes or tasks, starting from proven methods where they fit.
Custom AI & geospatial solutions
Purpose-built AI for specific imagery challenges.
When a standard workflow is not enough, Satvitech can scope tailored computer-vision models, automated processing pipelines, and specialist tools around the available data and the customer's objective.
Workflow
Ingest imagery
Preprocess
Run model
Review results
Export
Illustrative interfaceT-01Storage tankAccepted
C-02Container stackAccepted
C-03Container stackNeeds review
What can be developed
Built around one task, done well
Every custom development relates directly to imagery, computer vision, image analytics or the geospatial workflows around them.
Automated analysis pipelines
Connect image ingestion, preprocessing, model inference, validation and output generation into one repeatable flow.
Repeatable batch processing
Organise repeat image-analysis tasks where the input data and its licensing allow.
Custom reporting & interfaces
Targeted review screens, dashboards or reports for a defined workflow — not general-purpose software.
Data export & integration
Structure results for your existing workflow once the formats and integration requirements are known.
Model evaluation & refinement
Test an approach against suitable reference examples, document limitations and improve it when the data supports it.
How custom projects begin
Discovery before development
A proof of concept with agreed success criteria comes before any larger build, so you can see what works on your data first.
-
Step 1: Describe the task
You describe the task and the decision or output you need.
-
Step 2: Review sample data
We review sample data, permissions, quality and feasibility.
-
Step 3: Agree a proof of concept
Together we agree a proof-of-concept scope and success criteria.
-
Step 4: Develop and test
A suitable model or workflow is developed and tested on representative data.
-
Step 5: Review and decide
Results and limitations are reviewed before production use or further development.
Modular by design
Pipelines that fit into your existing work
Custom workflows are built from clear stages, so each one can be tested, replaced or improved on its own. Outputs are shaped for the place they will be used — a report, a spreadsheet, a GIS layer or another system.
- InputsLicensed, open or customer imagery
- ProcessingPreparation, model inference, rules
- ReviewHuman checks where it matters
- OutputsFiles, reports, review screens, exports
Flexible engagement
Start small, grow only if it works
Scope and pricing depend on the imagery, requirements and delivery approach. We agree them with you once the task is understood.
Feasibility assessment or proof of concept
Test whether the task is achievable with the available data before a larger commitment.
Defined analysis project
A scoped piece of work with agreed inputs, method and deliverables.
Custom workflow development
A repeatable pipeline or tool built for a specific, recurring task.
Ongoing refinement or processing
Continued improvement or recurring processing, where mutually agreed.
FAQ
Custom development questions
Do you build general software or websites?
No. Custom work is limited to AI, imagery, computer vision, image analytics and the geospatial workflows around them — for example a model, a processing pipeline or a review interface for analysis results.
Do we need training data?
It depends on the task. Some problems can use existing methods with little adaptation; others need labelled examples. We assess this during the feasibility stage and explain what would be needed.
How long does a custom project take?
It depends on the data, the complexity of the task and the agreed scope. We propose a timeline after reviewing sample data rather than quoting one in advance.
Related services
Combine with other capabilities
Custom solutions often extend one of our standard services to a new class, data source or output format.
Tell us what you need to analyse
Describe the task, the decision it supports and any sample imagery you have. We will suggest whether a proof of concept is the right first step.