With funding from the government, we have been researching the topic of deep learning for quite some time.
A brief and simplified explanation:
Deep-learning is about teaching an artificial neural network structure to make its “own” decisions, to control processes and to carry out classifications.
Our goal at AKL-tec is to use deep learning to develop a method that makes it possible to use as few sensors and cameras as possible to record product data and still generate highly precise data.
Deep-learning consists to a large extent of teaching the machine, which is often done manually by classifying real images of the customer’s environment. In our case, this involved processing well over 10,000 images to teach a camera (with PC) to recognize and classify pallet labels (IPPC, CHEP, EURO labels, etc.) fully automatically.
This work was done in cooperation with our development partner. Consequently, each teach-in results in a highly specialized, customer-oriented solution, which in most cases can only be transferred to other customer requirements to a limited extent. Thus, machines ideally have to be retrained for each project in order to achieve maximum precision.
We at AKL-tec GmbH are enthusiastic about such and other projects and are looking forward to development requests!
