Anomaly detection
Learn what normal looks like and flag anything unusual — useful when faults are rare or hard to describe in advance.
Back to computer vision
IMAGE MACHINE LEARNING · POSSIBILITIES
What your images
could be doing.
Most businesses already capture images — from cameras, phones, scanners and product shoots. Image machine learning turns them into counts, checks, alerts and data you can act on. Here are some of the ways it can help.
01 / IN PRACTICE
Six ways to put
your images to work.
Each example below is a common starting point. Real projects are shaped around your own images, systems and the decisions you want to support.
01 / QUALITY INSPECTION
Catch defects before they reach a customer.
A camera above a production line or packing bench checks every item against what “good” looks like. Scratches, cracks, missing parts or incorrect labels are flagged in milliseconds, so your team can focus on the items that need a closer look.
- Surface scratches, dents & cracks
- Missing components or incorrect assembly
- Label, print & packaging checks
Every item checked the same way, at line speed.
Quality inspectionILLUSTRATIVE DEMO
PASS 0.99 CRACK 0.94
INSPECTED TODAY4,812FLAGGED17PER CHECK21 ms
DEFECTS DIVERTED FOR HUMAN REVIEW 0.35%
02 / DETECTION & COUNTING
Count stock without the clipboard.
Object detection finds and counts individual items in a photo or video frame. A quick picture of a shelf, pallet or yard can become a stock count, with gaps and misplaced items highlighted automatically.
- Shelf & warehouse stock levels
- Pallet, parcel & vehicle counts
- Gap and layout checks
Faster counts, with the photo kept as the record.
Detection & countingILLUSTRATIVE DEMO
gapgap
DETECTEDSKU-104SKU-221SKU-318Gaps to restock
COUNT SYNCED TO STOCK SYSTEM 1.2 sec
03 / DOCUMENT EXTRACTION
Turn paperwork into usable data.
Vision models can read scanned and photographed documents, find the fields that matter and pass them into your systems. Invoices, delivery notes and forms no longer need typing in by hand, and anything the model is unsure about is sent to a person to check.
- Invoices & receipts
- Delivery notes & job sheets
- Handwritten forms & certificates
Less manual data entry and fewer typing errors.
Document extractionILLUSTRATIVE DEMO
INVOICEDale Timber Supplies LtdNo. INV-2041812 Sep 2026Total £1,284.60PO 5521
- supplier
- Dale Timber Supplies Ltd0.99
- invoice_no
- INV-204180.99
- date
- 2026-09-120.98
- total
- £1,284.600.97
- po_ref
- 5521check
4 FIELDS FILLED · 1 SENT FOR REVIEW 0.8 sec
04 / CLASSIFICATION & TAGGING
Sort and tag images automatically.
Image classification assigns each picture to a category and adds descriptive tags. Product catalogues, photo libraries or images sent in by customers can be organised consistently and made searchable, without anyone labelling them one by one.
- Product catalogue attributes
- Routing customer-submitted photos
- Organising archives & image libraries
Consistent labels across thousands of images.
Classification & taggingILLUSTRATIVE DEMO
Clothing+1Furniture+1Lighting+1
TAGGED + FILED AUTOMATICALLY 0.3 sec / image
05 / SAFETY & SITE MONITORING
Spot hazards as they happen.
Existing cameras can watch for the situations that matter: someone entering an exclusion zone, missing protective equipment or a blocked fire exit. Alerts reach the right person straight away, and the system can be designed to detect people without identifying them.
- Exclusion zones around machinery
- PPE & hi-vis checks
- Blocked exits & walkways
Earlier warnings, designed with privacy in mind.
Safety & site monitoringILLUSTRATIVE DEMO
EXCLUSION ZONEforkliftperson · PPE ✓person · PPE ✓IN ZONE
- Person entered exclusion zone — supervisor alerted
- Fire exit B clear
- Hi-vis check · 12 of 12 people
ALERT SENT · NO FACES STORED 0.4 sec
06 / VISUAL SEARCH
Find it from a photo.
Visual search compares what images look like rather than the words attached to them. Customers can photograph something and find similar products, and staff can match a spare part, a pattern or a previous job in seconds.
- “Find similar” for online shops
- Spare part identification
- Spotting duplicates & near matches
Match by appearance when words fall short.
Visual searchILLUSTRATIVE DEMO
QUERY PHOTOVISUAL FINGERPRINT
- Field watch 380.94
- Field watch 400.91
- Pilot watch0.86
- Dress watch0.79
RANKED BY VISUAL SIMILARITY 140 ms
02 / AND MORE
Other places
vision can help.
Measurement & segmentation
Outline regions pixel by pixel to measure sizes, areas, coverage or damage directly from an image.
Drone & aerial imagery
Survey roofs, land, crops or assets across large areas and track how they change over time.
Agriculture & produce
Grade produce, estimate yields or spot signs of disease and pests from photos taken in the field.
Image enhancement
Clean up, crop and standardise product or property photos automatically before they are published.
Privacy protection
Blur faces and number plates automatically before footage or photos are stored or shared.
03 / GETTING STARTED
What a good project
needs to succeed.
Representative images
Examples that reflect real conditions — the lighting, angles and variety you actually see. It is often fewer than you might expect to test an idea.
A clear measure of success
Agree what accuracy is useful and what a mistake costs. A missed defect and a false alarm rarely matter equally.
The right place to run it
On a small device beside the camera, on your own server or in the cloud — chosen for speed, cost and where your data needs to stay.
People in the loop
Uncertain results go to a person to review, and their decisions can be used to improve the model over time.
04 / YOUR IMAGES, PUT TO WORK
Have an idea
for your images?
Tell me what you capture and what you would like to know from it. We can talk through whether image ML is a good fit and what a small first test could look like.
Open to projects in North Yorkshire and nationwide