# CPU or GPU for computer vision in egg inspection?

> Classical computer vision, such as colour, morphology and watershed, runs well on a CPU. A GPU makes the biggest difference for neural networks, which depend on many parallel multiplications. Vigiovo does not use neural networks and runs on an ordinary CPU, at about 90 frames per second on one core.

Updated: 2026-10-09 · URL: https://vigiovo.com/en/resources/cpu-vs-gpu-for-egg-inspection · Português: https://vigiovo.com/recursos/cpu-ou-gpu-para-inspecao.md

## Key takeaways

- Classical vision, such as colour and watershed, runs well on a CPU, and Vigiovo runs on one CPU core, without a GPU.
- A GPU gains the most with neural networks, which Vigiovo does not use.
- A GPU adds purchase cost, power draw, heat and driver maintenance.
- Measure throughput with your own images before deciding on hardware.

## Which processor does camera egg inspection need?

For classical vision, an ordinary CPU is usually enough. Colour segmentation, the distance transform and watershed are fixed rules applied to the image, and they run well on a CPU.

A GPU makes more difference for neural networks, because these networks perform many matrix multiplications that benefit from parallel processing. Those who choose neural networks usually need a GPU to keep up with line speed. Those who choose deterministic vision, as [Vigiovo](https://vigiovo.com/en/how-it-works) does, usually do not. See [deterministic vision vs neural networks](https://vigiovo.com/en/resources/deterministic-vision-vs-neural-networks) for the comparison of the two methods.

## What are the practical differences between a CPU and a GPU?

The table compares what usually weighs on a line's decision. It gives no prices and no laboratory measurements. It is a practical judgement, meant to guide the conversation with your IT team.

**Practical, qualitative judgement, no prices**

| Criterion | Ordinary CPU | Dedicated GPU |
| --- | --- | --- |
| Classical vision, such as colour and watershed | Suitable; this is where Vigiovo runs | Can be used, but not needed |
| Neural networks | Possible, but usually slow for real time with large models | The usual route to keep up the speed |
| Purchase cost | Lower, and available in any ordinary computer | Higher, since it is an extra component |
| Maintenance | Operating system and routine updates | Specific drivers, heat and extra power draw |
| Replacement | Simple, with common spare equipment | Depends on compatibility with the computer and power supply |

Each option has its costs. Before deciding, consider the hardware you already have, who will maintain it and the throughput the line really needs.

## How much processing does Vigiovo need?

Vigiovo runs on a CPU, without a GPU. In the project's tests, with 960×540 video and about 45 eggs on screen, it processed about 90 frames per second on a single core, with a mean latency of about 11 ms per frame. A single still image with 54 eggs was measured in about 41 ms.

These numbers come from a benchmark run by the project on one CPU core, and from synthetic scenes with known answers. They show the headroom of an ordinary processor, but they do not guarantee the same result on your computer. To work out the numbers for your line speed, see [latency and throughput in inspection](https://vigiovo.com/en/resources/latency-and-throughput-in-inspection).

## When does a GPU make sense?

A GPU makes sense when the system uses neural networks, or when the required throughput cannot be reached on an ordinary CPU even after optimising the processing. Before buying, confirm with measurements: if the throughput you need is already reached on a CPU, a GPU becomes cost without gain.

Vigiovo does not need a GPU to grade eggs. It records the reason for each egg in a [per-egg record](https://vigiovo.com/en/resources/per-egg-records-for-audits), so an audit depends on the measurements stored, not on the processor used.

## Does a faster processor solve the camera's limits?

No. A faster processor does not let the camera see inside the egg, and it does not reduce the effect of glare or uneven light. The limits of light and position apply whatever the hardware.

For the other side of the equation, see [how to choose a camera for egg inspection](https://vigiovo.com/en/resources/choosing-a-camera-for-egg-inspection).

## Frequently asked questions

### Can an ordinary CPU grade eggs in real time?

For classical vision, yes. In Vigiovo's tests, running on a single CPU core, the system processed about 90 frames per second with about 45 eggs on screen. Speed still depends on your camera and your line, so the test should be run with your own images.

### Do I need a GPU to use deterministic vision?

No. Deterministic vision uses classical image operations, such as colour segmentation, the distance transform and watershed, which run on an ordinary CPU. A GPU is usually associated with neural networks, which Vigiovo does not use.

### Does a GPU reduce the cost of inspection?

For classical vision, usually not. A GPU adds purchase cost, power consumption, heat and driver maintenance. It only pays off when the work needs acceleration that an ordinary CPU cannot deliver. Compare the total cost with the throughput your line actually needs.

## Sources

- [Como o Vigiovo classifica ovos sem IA / How Vigiovo grades eggs without AI](https://vigiovo.com/como-funciona), Vigiovo

## Read next

- [Latency and throughput: how many frames per second you need](https://vigiovo.com/en/resources/latency-and-throughput-in-inspection.md) (Explainer, 2026-10-09): Latency and throughput in video egg inspection: how long one frame takes in Vigiovo, how many frames per second it processes and how to work out what your line needs.
- [Per-egg records for grading audits](https://vigiovo.com/en/resources/per-egg-records-for-audits.md) (Guide, 2026-10-09): How a per-egg record, with class, scores, reasons and measurements, supports grading audits and the equipment verification that MAPA asks for.
- [Candling vs computer vision: what each one detects](https://vigiovo.com/en/resources/candling-vs-computer-vision.md) (Comparison, 2026-10-09): Candling compared with camera inspection in egg grading: which defects each detects, speed, record keeping and why the two methods work best together.

## How Vigiovo helps

Vigiovo grades eggs on the conveyor as clean, dirty, cracked or unverified, in real time, with deterministic computer vision: every decision comes with the measurements behind it, on an ordinary CPU. https://vigiovo.com/en/how-it-works.md

3 min read.