Framework for the integration of deep learning into intelligent systems
Advanced machine learning algorithms, including Convolutional Neural Networks (CNN) or Deep Learning, have revolutionised robotics over the last decade. Especially the perception, i.e. the ability of robots to understand their environment by means of sensor systems, is affected. An important example in this context is 3D computer vision and 3D image understanding, which refers to the analysis and recognition of objects via volumetric images and point clouds.
The lack of standardised industrial data analysis platforms has made it difficult to use deep learning modules in industrial robot systems. However, it is of great importance to have such a platform that can coordinate different Deep Learning modules in a complex system and make the output of one data processing module available to all other robot modules in real time.
The aim of the project is to develop a framework that integrates deep learning into distributed intelligent systems. Fields of application are industrial robotics and 3D computer vision, whereby three concrete use cases of industrial automation are considered. The result will be a faster and more accurate image analysis and a better understanding of the image without the need for hardware upgrades. This makes it possible to use it in both existing and new environments in a wide range of sectors.
Project: Deep-Learning in Robotics and 3D Computer Vision (its-3DL)
Project duration: Spring 2020 to spring 2023
Project volume: EUR 1.4 million