Computer vision
also: machine vision · vision inspection · image recognition
Using cameras and algorithms, increasingly neural networks, to detect, measure, classify or inspect objects, from factory checks to products that see.
In manufacturing, machine vision inspects parts for defects, reads labels and codes, measures dimensions and guides robots. In products, vision enables occupancy counting, safety monitoring, wildlife cameras and smart appliances. Modern approaches use neural networks trained on labeled images; classical image processing still works well for controlled, repeatable scenes.
Results depend heavily on optics and lighting, which often matter more than the algorithm: consistent, purpose-designed lighting turns a hard vision problem into an easy one.
Where it bites: collecting enough labeled examples of rare defects, changes in lighting or product appearance over time, privacy rules for cameras that capture people, and compute and power for on-device processing.
Founder takeaway
Design the lighting and camera placement first, collect images from the real environment, and plan how the model will be updated as conditions change.
engineer review pending