What does our image analysis platform offer?

We provide image recognition software and cloud infrastructure for data analysis and storage.

Our platform is used by hundreds of customers worldwide and is also available for partners to host or integrate their own image analysis pipeline. This can be done via our API for existing imaging solutions or with a new custom solution.

What does our image analysis platform offer?

We provide image recognition software and cloud infrastructure for data analysis and storage.

Our platform is used by hundreds of customers worldwide and is also available for partners to host or integrate their own image analysis pipeline. This can be done with existing imaging solutions or brand new hardware. Via our API you can also integrate any existing software already in use into our infrastructure.

Existing Products

FW: Fermentation Wine

A smart, technology driven solution for automated yeast analysis that combines a 400x optical magnification with cloud-based image analysis software. Specifically designed Specifically designed for wine makers and vintners.

BB: Better Brewing

An automated yeast analysis system specifically designed for brewers. Combines a mobile microscope with cloud-based image analysis software and gives accurate results 10 times faster than a traditional counting chamber. Product launched in 2016.

Products in Development

Water online monitoring

Development of an online monitoring system for the microbiological status of drinking water and industrial water.

MUH: Monitoring Uterine Health

A quick and reliable on-site diagnostic tool for detecting subclinical endometritis in dairy cows.
Based on automated image recognition and analysis, using deep learning and artificial intelligence.

Proven technical feasibility

Asbestos Fibre Detection
Asbestos Fibre Detection
Blood Count
Asbestos Fibre Detection
Parasite Egg Detection
Asbestos Fibre Detection

Metal Analysis
Asbestos Fibre Detection

Bitumen Emulsion Analysis
Asbestos Fibre Detection

Cytological Analysis
Asbestos Fibre Detection

  • In an ongoing R&D project with partners from industry and academia we are developing a system for the detection of bacteria in enriched water samples.

  • Oculyze Image analysis platform can be used to analyze materials such as metal structures

  • Our Technology can also be used for analysis of emulsions such as bitumen for road construction as shown here

  • Within an ongoing academic project we are supporting the adaption of our technology for the use of in blood counts as well as automated malaria detection

  • The Oculyze Image analysis platform is suitable for the detection of parasite eggs in stool samples

  • Successful detection of asbestos fibers in SEM (scanning electron microscope) images has been demonstrated

Modern Restful API available for existing products & also your tailored solutions

Sample Documentation

Why it makes sense to run your AI built computer vision software in the cloud

AI Computer Vision Software

Using machine learning and Artificial Intelligence (AI), Oculyze has transferred the computer vision software for lab equipment from the table top to the cloud. We automate expert image analysis combining methodical pattern recognition with artificial intelligence (AI) and deep learning to create some of the best computer vision software money can buy. This base technology, used in the Oculyze yeast cell counters, Better Brewing and Fermentation Wine, has convinced hundreds of yeast labs, breweries and wineries of all sizes around the world.

By AI computer vision software we mean software that is able to do useful things, but without all the instructions being hard coded. Traditional software needs all possible cases to be taken into consideration during the initial programming. Since we mostly automate the analysis of images with a lot of variance, noise and biological diversity this feature comes in very handy as the software performs great on scenarios none of us has ever seen before and does so consistently.

Our first algorithms were specifically trained to count yeast cells in very challenging situations (high concentrations, in clusters and mixed with other particles). By watching thousands of these images and counting the cells in them over and over again the algorithms learned what is a live cell, a dead cell and how many cells are actually in a specific cluster. The algorithms were helped by traditional image pre-processing taken from traditional pattern recognition techniques.

Critics of this so called narrow AI say that it is actually artificial experience (AE) and not intelligence. It is estimated that it took the deep learning network “five” the equivalent of 45.000 years to beat humans in the game Dota 2. While this shows how much “experience” went into the intelligence of this network it also makes it easy to understand why this type of software is so superior for many tasks. When our hardware was successfully validated by the @vlb in 2016, the algorithms had been in training for less than one human year, yet the system performed as well as a professional with 20 years of experience.

Hard coded image recognition devices have been around since the 1950’s and combined the worst of two worlds- they were as expensive as the yearly salary of a human expert and were not able to learn from the samples they analyzed. Unless their software is later manually re-programmed the software stays the same forever. The price for these devices has dropped significantly in the last decade but until recently they were not able to gain any experience and the software did not improve over time.

In the Cloud

This changed when we started to use the cloud for other things than streaming music and storing pictures. The cloud gave us the cost efficient ability to use extremely high computing power by proxy on affordable handheld devices (<200 EUR/USD). Our computationally demanding algorithms would not be possible to run locally in a practical timeframe, but because the calculations are being performed in the cloud you have it instantly available on a handheld device.

As a result

Real beauty happens when you combine the two components, AI and the cloud, and gain a system that is flexible and affordable, using the experience from the many for the benefit of each individual (user). The samples from the devices allow the algorithms in the cloud to keep learning and improving the computer vision software for all users. This is how Oculyze computer vision software keeps getting better and better.

Some claim that a AI network can’t apply what it learned while playing Chess to play Go, reinforcing the argument that it is actually artificial experience and not artificial intelligence. While that may be true for different games, we have noticed that the experience gathered while counting yeast is helping our algorithms to count fibers, cylinders and other shapes better and quicker. This allows us to dramatically reduce the amount of images we need in our image analysis platform to train the initial algorithm for new applications.

In a time of shortage of skilled workers it makes a lot of sense to automate visual analysis tasks and reduce the reliance on human experience in favor of an artificial intelligence that stays within the company. It does not take 45.000 years to teach a human how to perform a manual visual analysis but it is impossible for a human to gain or access the combined experience available to cloud based computer vision software.

Combining Results

Since you have all your data already saved in a standardized format in the cloud, it becomes possible to also use AI for more advanced interpretations of the data. For wineries for example we help predict problems with fermentation by combining the results of various measurements, thus allowing our customers to react earlier and avoid problems before they even occur.

In summary: Cloud based, AI trained computer vision software is faster to develop and deploy, cheaper, more accurate and keeps improving over time.

What analysis are you thinking about automating? Send us a mail and include a picture of the sample.

Backed by years of research

Automatisierte Klassifizierung und Viabilitätsanalyse von Phytoplankton


A simple viability analysis for unicellular cyanobacteria using a new autofluorescence assay, automated microscopy, and ImageJ


PlanktoVision – an automated analysis system for the identification of phytoplankton


The use of fluorescence microscopy and image analysis for rapid detection of non-producing revertant cells of Synechocystis sp. PCC6803 and Synechococcus sp. PCC7002

- Katja Schulze, Imke Lang, Heike Enke, Diana Grohme, and Marcus Frohme


FIJI Macro 3D ART VeSElecT: 3D Automated Reconstruction Tool for Vesicle Structures of Electron Tomograms

- Kristin Verena Kaltdorf, Katja Schulze, Frederik Helmprobst, Philip Kollmannsberger, Thomas Dandekar , Christian Stigloher