Embedded Vision Systems Group

The Embedded Vision Systems (EVS) group works within the Computer Vision Laboratory, Department of Automatic Control and Robotics, Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering (EAIiIB), AGH University of Krakow, Poland.

We are involved in research on implementation of various types of vision systems, which can operate in real time even for high-resolution video streams and sampling – up to UHD (3840 x 2160 @ 60 fps) – while maintaining low power consumption. As hardware platforms, we use reprogrammable FPGAs (Field Programmable Gate Array), reprogrammable SoCs (System on Chip) such as Zynq SoC, Zynq UltraScale+ MPSoC (Multi Processor System on Chip), or embedded GPU (Graphic Processing Unit) solutions such as Nvidia’s Jetson series. We are also interested in neuromorphic platforms.

Such systems are used in applications requiring very fast and efficient processing of video data so that the information obtained from them can be used almost immediately to support decision-making or control. Current research focuses on the following areas:

• control of autonomous, unmanned flying vehicles (drones, UAVs, UAS),
• vision algorithms for autonomous vehicles (self-driving cars) and Advanced Driver Assistance Systems (ADAS),
• object tracking and trajectory determination,
• use of LiDAR data for object detection and environment mapping,
• segmentation of foreground and moving objects, as well as elements of advanced video surveillance systems (AVSS – Advanced Video Surveillance Systems),
• implementation of embedded AI methods - deep convolutional neural networks and spiking neural networks.

Since 2011, members of the group have been co-organizers of the international conference DASIP (The Conference on Design and Architectures for Signal and Image Processing), cooperating with scientists from Germany, France, Italy, Spain, Portugal and Canada. We also present our research at the ARC (International Symposium on Applied Reconfigurable Computing), ICCVG (International Conference on Computer Vision and Graphics), SPA (Signal Processing Algorithms, Architectures, Arrangements and Applications), KKA (National Conference on Automation), FedCSIS (Federated Conference on Computer Science and Information Systems) conferences and in Journal of Real-Time Image Processing, Journal of Signal Processing and Sensors.

The team members teach in the following courses: Digital Image Processing and Vision Systems (Polish and English), Reconfigurable Systems, Structured and Object-Oriented Programming, Advanced Object-Oriented Programming, Algorithms and Data Structures, Architectures of Vision Systems, Image Compression and Coding Methods, Software-Hardware Implementation of Algorithms (Hardware-Software Vision Systems), Perception Systems and Algorithms in Autonomous Vehicles, Parallel Computing in CUDA, Biometrics, and HLS tools. Every year, several bachelor and master theses are defended as a result of cooperation between staff and students. In addition, our team closely cooperates with the AGH AVADER Student Research Group, which brings together students interested in embedded vision systems, drones and autonomous vehicles.