Edge Vision Inspection
An experimental prototype for detecting visual defects using low-cost cameras.
Overview
Manual visual inspection is tiring and inconsistent, while industrial vision systems are often too expensive for smaller producers.
This experimental prototype explores whether low-cost cameras running AI on the device itself can flag visible defects in real time.
The challenge
Defects are missed late in the shift, inspection results aren't recorded, and commercial systems are out of budget for many factories.
Our approach
- 01
Sample
Collect images of good and defective parts in real lighting.
- 02
Train
Train a compact detection model suited to small devices.
- 03
Deploy
Run the model on an edge device next to the line.
- 04
Field-test
Compare results with human inspectors and refine.
Key features
On-device inference
Runs without a constant internet connection.
Defect heatmaps
Highlights where on a part the issue was detected.
Operator interface
Simple pass / check screen for line staff.
Inspection log
Every result stored for quality reporting.
Designed to
- Test affordable automated inspection
- Record inspection results consistently
- Support, not replace, human inspectors
These are design goals for a concept, not measured results.
More projects
Multilingual Knowledge Assistant
A concept assistant that answers questions from an organisation's documents in Dari, Pashto and English.
LLMs · Vector search · TypeScript
Operations Analytics Hub
A concept dashboard unifying operational data with forecasting for faster decisions.
Python · PostgreSQL · React
Document Intake Automation
A concept pipeline that reads scanned forms, extracts fields and routes them for approval.
Computer vision · Workflow engine · APIs
Have an Ambitious Idea?
Let's explore how intelligent technology can help bring it to life.
Start a conversation