Rocky Labs
AI features

Computer vision

Software that reads images and video to count, inspect, scan and measure.

We build vision features that run on a phone, in the browser or on a server. That includes reading documents and labels, detecting and counting objects, and checking photos against a standard. Where it helps privacy or speed, the model runs on the device and images never leave it.

This is for you if

  • Someone checks photos or video by eye, and it's slow or inconsistent.
  • You need to read text from labels, receipts or forms out in the field.
  • You want your app to recognise what the camera is pointed at.

What you get

  • A feasibility test on your real images
  • A labelling plan, and the labelling itself if needed
  • A trained or fine-tuned model, or a vision API integration
  • On-device or server deployment
  • An accuracy report on images the model hasn't seen
  • Integration into your app or workflow

What this can look like

Examples of the kind of work we take on.

Document and label scanning

Read text, barcodes and fields from receipts, labels or forms, on the device.

Visual inspection

Flag defects or missing items in photos from a production line or a site visit.

Counting and measuring

Count items in a photo, or estimate size using a reference object in the frame.

Tools we use for this

  • Apple Vision
  • Core ML
  • Google ML Kit
  • OpenCV
  • YOLO
  • PyTorch
  • Claude and OpenAI vision models

Common questions

How much data do we need?

Often less than people expect. Many tasks work with a few hundred labelled examples. We run a small feasibility test first, so you know before paying for a full build.

Can it work offline?

Often, yes. Many models run well on a recent phone, which also keeps images private.

For anything else, email [email protected].

Related services

Got an app idea or a process you'd like to automate? Tell us about it.Start a project