Artificial intelligence in embedded systems
Artificial intelligence has found its way into many areas to make systems more precise and reliable than previous manually optimized algorithms. One common method is machine learning, which learns to solve predefined tasks on its own based on a large database. But for artificial intelligence to enrich the end application, it must be embedded. A particular challenge here is to compress the networks so that they can be completely computed in the end devices.
Benchmarking for AI Accelerators
When using artificial intelligence in practice, different framework conditions have to be considered that influence the design of an AI system and accelerator. Often, a trade-off is realized between the accuracy of the AI model, hardware resources used, and application requirements. At ITIV, metrics from the different domains are measured to analyze this trade-off and benchmarks are developed to evaluate the overall design. These benchmarks can in turn be used in automated design space exploration to find a suitable hardware/software co-design more quickly.
Many people strive for a long, self-determined life within their own four walls. However, the care crisis means that human care in the home is becoming increasingly rare. As an alternative, KIT is researching assistive robots that help elderly people to cope with everyday life at home. Important aspects for increasing their acceptance are short response times and personalization of the services. For this purpose, ITIV is researching suitable hardware accelerators with low latencies and a high degree of privacy.
|Design and implementation of a Deep Neural Network hardware accelerator for face recognition on FPGA.||Bachelor-/ Masterarbeit||ab 03 / 2023|
|Development of a hardware accelerator for Graph Neural Networks for object recognition in embedded systems||Master thesis||ab 03 / 2023|
|Implementation of a hardware accelerator for neural networks for processing radar data||Masterarbeit||ab 02 / 2023|
|Methodology for evaluating model-accelerator co-design in deep learning||Masterarbeit||ab 08 / 2023|
|Integration of hardware accelerators in humanoid robot platform||
ab 08 / 2023
Supervised student work (selection)
- MA: "Konzeptioneller Entwurf eines modularen Sensornetzwerks für intelligente Textilanwendungen"
- MA: "Datenanalyse von Sensorinformationen in intelligenten Textilanwendungen"
- MA: "Structured Analysis of a Deep Neural Network for Face detection for Implementation on FPGAs"
- MA: " Design and Analysis of an Intelligent Sensor Network for Motion Tracking"
- MA: "Design and Analysis of a Human Pose Estimation System from Sparse IMU-Sensing"
- MA: "Efficient Design of 3D-CNN-Acceleration on FPGA for Action Recognition"
- BA: "Analysis of concepts for an AI-based system for automated identification and assignment of machine parameters"
- SA: "IMU-based Action Recognition using Machine Learning"
Kreß, F.; Sidorenko, V.; Schmidt, P.; Hoefer, J.; Hotfilter, T.; Walter, I.; Harbaum, T.; Becker, J.
2023. Computer Networks, 229, Article no: 109759. doi:10.1016/j.comnet.2023.109759
Kreß, F.; Hoefer, J.; Hotfilter, T.; Walter, I.; El Annabi, E. M.; Harbaum, T.; Becker, J.
2023. Machine Learning and Principles and Practice of Knowledge Discovery in Databases. Hrsg.: I. Koprinska. Pt. 1, 557–568, Springer International Publishing. doi:10.1007/978-3-031-23618-1_37
Kreß, F.; Hoefer, J.; Hotfilter, T.; Walter, I.; Sidorenko, V.; Harbaum, T.; Becker, J.
2022. 18th International Conference on Distributed Computing in Sensor Systems (DCOSS), 133–140, IEEEXplore. doi:10.1109/DCOSS54816.2022.00034
Walter, I.; Ney, J.; Hotfilter, T.; Rybalkin, V.; Hoefer, J.; Wehn, N.; Becker, J.
2022. Machine Learning and Principles and Practice of Knowledge Discovery in Databases – International Workshops of ECML PKDD 2021, Virtual Event, September 13-17, 2021, Proceedings, Part I. Ed.: M. Kamp, 339–350, Springer International Publishing. doi:10.1007/978-3-030-93736-2_26