# How many frames per second do you need to inspect eggs in real time? > Latency is the time Vigiovo takes to process one video frame: about 11 ms on average, and 15 ms at p95 (95% of frames). Throughput is how many frames per second it processes, about 90 on one CPU core at 960×540. What your line needs depends on the conveyor speed and on how many eggs appear at the same time. Updated: 2026-10-09 · URL: https://vigiovo.com/en/resources/latency-and-throughput-in-inspection · Português: https://vigiovo.com/recursos/latencia-e-vazao-na-inspecao.md ## Key takeaways - Latency is the time for one frame; throughput is how many frames per second the system processes. - Measured on one CPU core, without a GPU: about 11 ms per frame on average and about 90 frames per second at 960×540. - Segmentation runs on every frame; defect measurement, the expensive part, runs only three times per egg. - There is no single number: the throughput you need depends on conveyor speed and on how many eggs are visible. ## What is the latency of one frame in Vigiovo? On video, Vigiovo takes about 11 ms on average to process a 960×540 frame, with about 45 eggs on screen. At p95, meaning for 95% of frames, the time is up to 15 ms. These values were measured on one CPU core, without a GPU, using the project's own benchmark commands. Latency matters for the decision, because each egg is only graded after its three checks. It measures the time of one frame from the start to the end of processing. ## What is the difference between throughput and latency? Throughput is how many frames per second the system can process. Latency is how long one frame takes to be processed. A system can have good throughput and still decide each egg late, if frames pile up in a queue. In Vigiovo the two are directly connected: with about 11 ms per frame on average, one CPU core delivers about 90 frames per second. Not every frame needs the same work, as the next section explains. ## How many frames per second does your line need? There is no single number, because the calculation depends on two factors in your line: the conveyor speed and how many eggs appear in the image at the same time. With a faster conveyor, each egg stays visible for less time and needs more frames to receive its three checks. So work it out with real data from your line, not with a generic value. The project's measurements serve as a reference for that calculation, and the [how it works](https://vigiovo.com/en/how-it-works) page explains the method that produces them. ## Why does segmentation run on every frame but measurement does not? [Colour segmentation](https://vigiovo.com/en/resources/glossary#colour-segmentation), which separates shell from background, is cheap and runs on every frame. Defect measurement is the expensive part: fitting each egg's [shading surface](https://vigiovo.com/en/resources/glossary#shading-surface) and measuring dirt and cracks takes more time. So each egg is measured only three times, not in every frame in which it appears. This split keeps latency low without giving up the checks. The rule that combines the three measurements is described in the article on [two-of-three voting](https://vigiovo.com/en/resources/two-of-three-voting). ## How do you work out the throughput your line needs? Do the calculation before installing, with data from your own conveyor. The steps below give an estimate to compare with the frame rate measured on the equipment. 1. Measure the real conveyor speed, in centimetres per second, in the stretch where the camera sees the eggs. 2. Record a video clip of the line with production lighting and with the normal density of eggs. 3. Check the frame rate the equipment delivers on that clip, with the system running. 4. Compare the distance an egg travels between two frames with half a radius of the egg, the minimum spacing of the checks. ## Frequently asked questions ### What is the latency of Vigiovo? On video, about 11 ms on average per frame, and 15 ms at p95, for 960×540 with about 45 eggs on screen, on one CPU core without a GPU. These values come from the project's benchmark and do not replace a test with video from your own line. ### How many frames per second are needed to inspect eggs? It depends on the conveyor speed and on how many eggs appear in the image at the same time. With a faster conveyor, each egg stays visible for less time and needs more frames for its three checks. Work it out with data from your line, not with a generic number. ### Are throughput and latency the same thing? No. Throughput is how many frames per second the system processes; latency is how long one frame takes to be processed. The two are related, but a system can have good throughput and still decide each egg late, if a queue builds up. ## Sources - [Como o Vigiovo classifica ovos sem IA / How Vigiovo grades eggs without AI](https://vigiovo.com/como-funciona), Vigiovo ## Read next - [Two-of-three voting: why measure each egg three times](https://vigiovo.com/en/resources/two-of-three-voting.md) (Explainer, 2026-10-09): How Vigiovo measures each egg in three checks as it passes on the conveyor and rejects it only when the defect shows up in two, which filters out isolated glare. - [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 4 min read.