Members of the Pervasive Computing group presented four papers at the 2024 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2024) in Seoul, Korea and the International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT 2024) in Abu Dhabi, UAE:
Rockson Agyeman introduced his Multi-Human Parsing Network (MHParsNet) for resource-efficient multi-human parsing. The architecture of MHParsNet uses a Feature Pyramid Network (FPN) to recognize humans and their corresponding body parts at different scales and sizes. MHParsNet uses two dedicated segmentation branches to maintain global context: one for differentiating human instances and the other for differentiating body part instances. Lastly, MHParsNet adopts the single-stage design approach because it provides a better resource-accuracy tradeoff.
Kyriakos Lite presented a navigation framework that exploits location uncertainty for the motion control of the robot and aims for improving the mission performance while reducing location uncertainty. The framework enhances an artificial potential field (APF) controller by an adaptive information-seeking (IS) force towards areas with low spatial uncertainty.
Bernhard Rinner presented recent work of Rockson’s PhD research proposing a decentralized approach to text-based person re-identification (re-ID). This method distributes the re-identification task to camera nodes. Each node employs a resource-efficient re-ID model for inference on a small gallery of persons in its field of view. He also presented our comparison of occupancy detection techniques based on scalar environmental sensors such as temperature, humidity, and CO2. We investigate recent methods based on machine learning techniques and assess their performance and resource consumption in a typical indoor setting.
Publications
Rockson Agyeman and Bernhard Rinner. Real-time Multi-human Parsing on Embedded Devices. In Proceedings of the International Conference on Acoustics, Speech and Signal Processing (ICASSP). Seoul, Korea 2024.
Kyriakos Lite and Bernhard Rinner. Navigation for Autonomous Robots with Adaptive Information-Seeking. In Proceedings of the International Conference on Acoustics, Speech and Signal Processing, Workshop on Signal Processing for Autonomous Systems. Seoul, Korea 2024.
Rockson Agyeman and Bernhard Rinner. Resource-efficient Text-based Person Re-identification on Embedded Devices. In Proceedings of the International Conference on Distributed Computing in Smart Systems and the Internet of Things (DCOSS-IoT 2024). Abu Dhabi, UAE 2024.
Matteo Bovo, Rockson Agyeman, Muhammad Arif and Bernhard Rinner. Evaluation of Occupancy Detection with Distributed Environmental Sensors for IoT Applications. In Proceedings of the International Workshop on Intelligent Systems for the Internet of Things (co-located with DCOSS-IoT 2024). Abu Dhabi, UAE 2024.