Skip to content

Infinity Neural · since 2011

The same camera that today only records can start measuring your operation.

It understands what happens in front of each camera and turns it into an alert or a number. Without changing the cameras and without touching the system you already have.

The expensive part isn't recording. The expensive part is watching.

A control room with forty screens does not have forty pairs of eyes. It has two. The rest of the video is recorded, stored and watched by nobody.

And when something happens, the really expensive part starts: someone sits down and searches backwards, minute by minute, until they find the moment. That time never appears on an invoice, but it gets paid for.

Counting is worse still. How many people came in, how many lorries unloaded, how long the line was stopped. It is counted by hand, written down by hand, and each day's number depends on who wrote it down.

The same camera that today only records can start measuring your operation.

You don't have to change the cameras. The one already up there stays where it is: what changes is what happens to its image.

Everything goes in. Only what matters comes out. Routine is let through and what matters becomes an alert with its time, its camera and its video — or a number that comes out the same every night.

Cameras that alert you

They watch an area and speak up when something happens that shouldn't: someone in the manoeuvring area without a high-visibility vest. What reaches you is an alert with its time, its camera and its video.

Cameras that count

They measure and add up over a line or an area drawn on the image. What reaches you is a number that comes out the same every night and can be audited: how many people crossed that door on Tuesday at seven.

Asking is for finding. Counting is for measuring. And they are not done the same way.

What it does, in one sentence.

A camera records. What it records is watched by someone, or by nobody. Infinity Neural sits in between: it looks at the image, understands what is there and what is happening, and lets the routine through.

What isn't routine comes out in one of two ways. Either it is an alert —with its time, its camera and its video— or it is a number that adds up and can be compared across shifts, weeks and seasons.

It doesn't replace anyone. It stops people spending their attention watching what isn't happening, which is almost all the time on almost any camera.

Two ways to have it.

One tunes a single case down to the last detail. The other covers the whole installation. They don't compete: they do different things and they are almost always fitted together.

When somebody tells us 'I want the camera to alert me if this happens', there is almost never one right answer. There are two routes, and which one applies depends on how many cases have to be covered, how much precision each one needs and whether that installation has a line.

Bespoke. What it actually means.

It means somebody looks at your camera before promising anything. Where it is mounted, what angle it has, what light there is at six in the evening and at six in the morning, what happens when it rains, what crosses in front of it and how often. That is where the first thing we say comes from, and sometimes it is that it can't be done — and that is bespoke too.

If it can be done, what gets tuned is the detection for THAT scene. It isn't picking a model out of a catalogue and switching it on: it is adjusting it to your product, to the way you stack it, to the clothes your people wear, to the floor markings in your industrial unit. That is why it reaches a precision a general system can't reach in that particular case: because it isn't solving every case, it is solving yours.

What it asks in return is two things. First, time: you have to record, watch what really happens on that camera for a few days and correct. Second, a connection: the video is processed off site, so that installation needs a line that can take it.

And a third that isn't a cost but is worth saying: each tuned case is one case. If what you want is forty different things across forty cameras, tuning them one by one doesn't add up for you or for us. That is when what you need is the other one.

IRIS Neural. What it actually means.

It means a system gets installed and from then on the whole installation is covered. Every camera, not one. Many cases at once, not one. And the video, the rules, the alerts and the history in one place, instead of spread across a recorder, an email and a spreadsheet.

The rules are written by your organisation, which is who knows what deserves an alert at four in the morning. IRIS automates applying them all the time without tiring, which is exactly what a person can't do in front of a wall of screens.

And it works with the internet cable unplugged. For a lot of people this is the first question and the last: there are installations that can't depend on a line, because of regulation, because they are isolated, or because the line that exists can't be guaranteed.

What it doesn't do is tune the way bespoke does. A system covering forty cameras can't be adjusted to each one's scene the way a detection built for a single camera would be. That isn't a limitation of IRIS: it is what covering a lot means.

And why the two are almost always fitted together.

Because a real installation doesn't have one problem: it has a background layer and two or three things that really matter. The background layer —knowing what is happening on every camera, having the video in order, having the obvious things flag up— is covered by IRIS. The two or three things that actually decide something are tuned bespoke.

One example of the ones that keep coming up: an industrial unit with forty cameras where the two things that hurt are that nobody walks in front of the forklift at the loading bay and that the counting at the door stands up to an inspection. IRIS covers the forty. Those two go bespoke, and one of them type-approved as well.

The odd thing is the opposite: paying for watchmaker precision in forty places where it isn't needed, or going without it exactly where it is.

How you choose.

  • A rare case, a critical one, or one that has to be audited → bespoke. There the precision pays for itself.
  • Many cameras and many cases at once → IRIS. Tuning forty things one by one doesn't add up.
  • No line, or one that can't be guaranteed → IRIS, which doesn't need one.
  • A number that goes into a report or an inspection → bespoke, with whichever type approval applies.
  • Almost any real installation → both. IRIS covering everything and bespoke where it matters.

And what we're not going to tell you.

That one is better than the other. It isn't. A system covering forty cameras can't be tuned like a detection built for a single one, and a detection built for a single one doesn't cover forty. That isn't anyone's limitation: it is how it works.

What we do say is which one is right for you, even if it is the cheap one. And if neither of them does what you want well, we say that too.

And this is what comes out.

So far that is how it is organised. What comes now is the only thing that really convinces: what you see on screen when it is running. Every one of these detections was recorded at a real installation, not simulated — and if yours looks like one of them, you already know what to expect.

Lo que Lemon Neural entiende

Personas: cuántas, por dónde y cuánto tiempo.

Lemon
Genera un heatmap de la cantidad de público en salon del comic
Lemon
Contabiliza personas por escaparates en centro comercial
Lemon
Conteo personas centro comercial
Lemon
Genera un heatmap del aforo
Lemon
Cuenta los pasajeros en el andén
Lemon
Conteo personas y zonas visitadas
Lemon
Conteo personas homologado CEM
Lemon
Mide el tiempo de cada vehiculo en la ventanilla
Lemon
Conteo personas y uso espacios restaurante
Lemon
Conteo personas y uso de mobiliario bar
Lemon
Monitorea aforos de parking
Lemon
Cuenta personas en cola y tiempo medio
Lemon
Mide el tiempo por cliente en caja
Lemon
Monitorea clientes en los pasillos y frente a estanterias
Lemon
Mide nivel de ocupacion stands

Cada una es una grabación del sistema funcionando. Pasa el ratón por encima —o tabula hasta ella— para verla en movimiento.

And running, in real places.

Casos grabados

Infinity Neural optimiza la ocupación y rotación de mesas en zonas de restauración

Infinity Neural mide ocupación y colas en centros comerciales con IA en tiempo real

¿Pierdes dinero por retrasos, colas y riesgos en el embarque? Infinity Neural para puertos y ferries

Análisis de flujos y zonas de interés en tiendas y supermercados con la IA de Infinity Neural

Revoluciona tu bar o discoteca con Lemon Neural

Lemon neural: optimizando la ocupación de carriles en piscinas

Control de aparcamientos con Lemon Neural

Transforma tu restaurante con Lemon y Rust Neural

Sin rodeos: Lemon Neural descifra cada movimiento humano

Counting people is easy. Making the number stand up isn't.

Everything above can be shown running. There is one more thing that can be shown signed by someone from outside — and that is usually the one that settles an official file.

Anyone can put a counter on a door. The hard part comes when someone asks where that number comes from. And they do ask. An inspection asks when an occupancy figure has to be justified. The landlord asks when the rent is calculated on footfall. The advertiser paying for impressions asks. The committee asks when it has to decide how much staff to put on a Saturday.

At that point, 'a system we have counts it' isn't enough. You need to be able to say who has checked that the system counts correctly.

See Neural Pax

What type approval is, and what it isn't.

Type approval doesn't mean the system never gets it wrong. It means an independent body has measured how much it gets wrong, under defined conditions, and has put it in writing with its stamp.

That is exactly what you need when the number leaves your building and goes into an official file: not a promise from the manufacturer, but the measurement of a third party that doesn't sell the product.

The body here is the Centro Español de Metrología, the one that in Spain makes sure a metre measures a metre.

The group's twelve certificates

Neural Pax.

Único
European type-approved counter for counting people autonomously, automatically and in real time.
CEM
Centro Español de Metrología. The type approval is a document and you can ask for it.
98–99 %
The precision recorded in that type approval. It is not a figure of ours: it is the one measured by whoever granted the approval.
3.072
People per second tracked at the same time at each entrance of the counting area.

What changes when the number is type-approved.

  • An occupancy figure can be justified to an inspection without arguing about the method.
  • A footfall-based rent is negotiated with a figure both sides accept.
  • A campaign is measured with a figure the advertiser doesn't have to take on trust.
  • This month's number is compared with last year's, because it is counted the same way.
  • And when somebody questions it, the answer is a document and not an explanation.

What we don't say.

That it is 'the most precise on the market'. We don't know: that would need every other one measured under the same conditions, and we haven't done it. What we do know, and can be checked, is that it is the only European one type-approved to do this.

Or that the type approval covers any mounting. It covers counting done the way the protocol says: with the camera where it has to be and with the field of view it demands. Mounted any other way it counts just as well or as badly as any other, but it is no longer type-approved — and that has to be said before, not after.

And what has to be decided.

That is everything: why it is needed, what it does, the two ways of having it, how it looks running and which part is measured by someone from outside. What is missing isn't something we can supply.

What is missing is looking at one of your cameras. Where it is, what angle it has, what can be seen from there at six in the evening and at six in the morning. That decides what can be got out of it, and there is no knowing before looking at it.

What people usually ask.

¿Puedo ajustar mis campañas de marketing a la demografía de mis clientes?

Sí, y hacerlo es clave para mejorar la efectividad de tus campañas. Cuanto más sepas sobre quién visita tu negocio (edad, género, si vienen solos o en grupo…), más fácil será diseñar mensajes, promociones o productos que realmente conecten con tu audiencia.

La mayoría de negocios físicos no tiene acceso a esta información sin recurrir a encuestas o programas de fidelización. Pero gracias a soluciones de análisis de video con inteligencia artificial como Lemon Neural , ahora es posible detectar y segmentar automáticamente el perfil demográfico de tus visitantes de forma anónima.

Esto te permite lanzar campañas dirigidas con mayor precisión, adaptar tus creatividades al público real que te visita y evaluar qué perfiles responden mejor a cada acción, todo en tiempo real y sin recopilar datos personales.

Leer la respuesta entera →
¿Cómo se mide el tiempo medio de atención al cliente en un mostrador?

Medir manualmente cuánto tiempo pasa cada cliente en un mostrador es ineficiente y propenso a errores. Sin embargo, existen soluciones basadas en inteligencia artificial que permiten realizar este cálculo de forma automática, precisa y en tiempo real.

Lemon Neural detecta:

Cuándo una persona llega al mostrador.

Leer la respuesta entera →
¿Existe alguna forma de medir las colas en las cajas para reducir la espera?

Sí, existen soluciones que permiten monitorizar de forma automática cuántas personas están haciendo cola y cuánto tiempo esperan. Esta información es clave para tomar decisiones rápidas, abrir nuevas cajas cuando sea necesario y mejorar la experiencia del cliente.

Lemon Neural analiza en tiempo real:

El número de personas en cola por caja .

Leer la respuesta entera →
¿Cómo optimizar el número de cajas abiertas según la afluencia?

Tener demasiadas cajas abiertas en momentos de baja afluencia desperdicia recursos. Tener pocas cuando el local se llena provoca largas esperas, mala experiencia y pérdidas. La clave está en anticiparse y tomar decisiones basadas en datos reales, no en intuiciones.

Lemon Neural te permite:

Medir en tiempo real cuántas personas hay en tienda.

Leer la respuesta entera →
¿Cómo controlar si las cajas de pago están saturadas?

Saber si las cajas están funcionando con fluidez es clave para evitar cuellos de botella, tiempos de espera excesivos y una mala experiencia de compra. Pero muchas veces el personal no tiene visibilidad inmediata de lo que ocurre en tiempo real.

Lemon Neural monitoriza continuamente las zonas de pago, detectando:

El número de personas en cola .

Leer la respuesta entera →
¿Qué puedo hacer para reducir las colas en mi tienda?

Las colas son uno de los principales puntos de fricción en la experiencia de cliente. Afectan la percepción del servicio, provocan abandono y afectan directamente a la rentabilidad. Reducirlas no siempre significa añadir más personal, sino actuar con datos.

Lemon Neural monitoriza en tiempo real la formación de colas , su longitud, tiempos de espera y velocidad de avance. Con esta información puedes:

Detectar patrones de saturación en ciertos horarios o zonas.

Leer la respuesta entera →
¿Cómo saber si mis clientes están esperando demasiado tiempo en cola?

Las colas largas generan frustración, abandonos y pérdidas de ventas. Pero muchas veces no se detectan a tiempo porque el equipo no tiene visibilidad en tiempo real, y tampoco es fácil medir cuánto espera cada cliente.

Lemon Neural te permite monitorizar automáticamente las colas : cuántas personas hay, cuánto tiempo lleva esperando cada una y si se forman cuellos de botella en momentos clave. Incluso puede alertarte cuando se supera un umbral de espera definido, para que tomes medidas inmediatas, como abrir más cajas o reforzar la atención.

Así, no solo evitas malas experiencias, sino que mejoras la eficiencia operativa y mantienes el flujo de clientes sin fricción.

Leer la respuesta entera →
¿Mi local es accesible para personas con movilidad reducida?

Tener rampas, ascensores o zonas amplias es fundamental, pero la verdadera accesibilidad no se mide solo por el diseño, sino por cómo se usa el espacio en la práctica. A veces, un local cumple con la normativa pero sigue presentando barreras reales: puertas estrechas, pasillos obstruidos o mobiliario mal distribuido.

Para saber si tu espacio es realmente accesible, necesitas observar cómo interactúan con él las personas con movilidad reducida: ¿pueden moverse sin obstáculos? ¿utilizan las rutas previstas? ¿hay zonas que evitan sistemáticamente?

Lemon Neural detecta de forma anónima la presencia de personas con sillas de ruedas u otros elementos de apoyo, y analiza sus recorridos dentro del local. Con esta información puedes identificar barreras no evidentes, validar la eficacia de tus rutas accesibles y adaptar el espacio para que sea verdaderamente inclusivo y funcional.

Leer la respuesta entera →

Verlo funcionando en IRIS Neural El sistema de vídeo del grupo, con su propia web.

Get started

Show us one of your cameras.

You tell us what you would like to know from one of your cameras and we tell you whether it can be done, with which family and what it would take. If it can't be done, we tell you that too.