Products

Aldebaran

Aldebaran is your ideal application for image recognition. It combines several artificial recognition methods and recognizes as many of your uploaded images as possible and saves them in keywords. Image information (EXIF) is also analyzed and categorized so you can search for it.

Antares

Just a quick sample about language recognition in times of artificial intelligence.

Atair

Just imagine you have some details about your products and do not want to write teasers for every product. With Altair you can!

Neura

Neura is our own development as a chatbot. It is based on our own artificial intelligence platform. This means that our solution does not need another provider in the background. With our methods of extracting content from your websites, we are able to optimally adapt our chatbot to your website.

Expertise

We have knowledge of various tools and programs that use neural networks. We can advise you to use the right tools for your plans in IT so that applications with cognitive power can turn your raw data into valuable information.

A neural network is a system of information technology that is based on human brain's structure and equips computers with features of artificial intelligence. Neural networks are characterized by the fact that computers can use them to solve problems independently and improve their skills. Whether or not you need human training to begin with depends on the artificial intelligence method used.

Neural networks are based on the structure of the human brain, which processes information via a network of neurons. Artificial neural networks can be described as models made up of at least two layers - an input and an output layer - and usually other layers in between (hidden layers). The more complex the problem to be solved by the artificial neural network, the more layers are required. A large number of specialized artificial neurons lie on each layer of the network.

Information processing in the neural network always follows the same sequence: information in the form of patterns or signals hits the neurons in the input layer, where it is processed. Each neuron is assigned a weight so that neurons are assigned different levels of importance. The weight, together with a transfer function, determines the input, which is now passed on. An activation function and a threshold value calculate and weight the output value of the neuron in the next step. Depending on the evaluation and weighting of the information, further neurons are linked and activated to a greater or lesser extent. This combination and weighting is used to model an algorithm that generates a result for each input. With each training, the weighting and thus the algorithm is adjusted so that the network delivers ever more accurate and better results.

Neural networks can be used for image recognition. Unlike humans, a computer cannot tell at a glance whether a picture shows a person, a plant or an object. He has to examine the photo for individual features. The computer knows which features are relevant from the implemented algorithm or it finds it out itself through data analysis. In each layer of the network the system checks the input signals, e.g. the images, based on individual criteria such as color, corners, shapes. With each test, the computer can better evaluate what can be seen in the picture. At first, the results will be relatively flawed. If the neural network receives feedback from a human trainer and can thus adapt its algorithm, it is called machine learning. With deep learning, human training can be omitted. In this case, the system learns from its own experience and gets better the more image material it has. In the end, ideally, there is an algorithm that can identify the content of the images without errors, depending on the training, regardless of whether these images are black and white or in which pose or from which perspective the depicted can be seen.

As an example we developed Aldebaran for your needs in image recognition.