---
title: "How to run a computer vision proof of concept"
description: "Start with one question a camera can answer, such as whether a part is defective or a space is occupied, and test it on images from your own cameras and conditions. Label a small, representative set, agree in advance the accuracy that would be useful, and measure the result on images the model has not seen. Include hard cases such as poor light, glare and unusual angles, and end with a decision to build, change or stop."
url: https://lidvizion.ai/answers/how-to-run-a-computer-vision-proof-of-concept/
site: Lid Vizion
index: https://lidvizion.ai/llms.txt
---

[Direct answers](https://lidvizion.ai/answers/)

# How should a company run a computer vision proof of concept?

Start with one question a camera can answer, such as whether a part is defective or a space is occupied, and test it on images from your own cameras and conditions. Label a small, representative set, agree in advance the accuracy that would be useful, and measure the result on images the model has not seen. Include hard cases such as poor light, glare and unusual angles, and end with a decision to build, change or stop.

## What this means in practice

- Pick one question a camera can actually answer.
- Use images from your own cameras, sites and lighting.
- Agree the accuracy that would be useful before you start.
- Test on images the model has not seen, including hard cases.
- Check privacy and where processing will run before you scale.

## How Lid Vizion approaches it

We start with the business decision and operating context, define ownership and measurable milestones, then use configuration, integrations, automation, or custom software only where each earns its place.

## More on Computer vision

- [Can computer vision monitor a workplace without identifying people?](https://lidvizion.ai/answers/can-computer-vision-work-without-identifying-people/)
- [What can computer vision do for a business?](https://lidvizion.ai/answers/what-can-computer-vision-do-for-a-business/)
- [How accurate does a computer vision model need to be before it is useful?](https://lidvizion.ai/answers/how-accurate-does-a-computer-vision-model-need-to-be/)

[Compare services](https://lidvizion.ai/services/) · [Use a planning tool](https://lidvizion.ai/tools/) · [Explore workflow use cases](https://lidvizion.ai/use-cases/) · [More direct answers](https://lidvizion.ai/answers/)

Talk it through

## Have a version of this question inside your company?

Bring the decision, the constraints, and what a good outcome looks like.

[Talk through the decision](https://lidvizion.ai/contact/?need=milestone-delivery&from=%2Fanswers%2Fhow-to-run-a-computer-vision-proof-of-concept%2F#write) · [Book a call](https://calendar.app.google/bB26LdA9e2gmGFp69)