Fares Abawi

Research Associate, University of Hamburg


Office: IKUM, F-214

Vogt-Kölln Str. 30

22527 Hamburg

Hi there! I’m Fares Abawi, a research associate and Ph.D. candidate at the University of Hamburg. My work is mostly about making computers and robots smarter and more perceptive, especially in understanding what grabs our attention in different situations. I’ve been particularly focused on projects like “GASP: Gated Attention for Saliency Prediction” and exploring how humanoid robots can mimic human-like attention and conflict resolution through multimodal integration.

As an advocate for collaborative technology, I actively contribute to open-source software projects (check out ImageBind-LoRA and Llama LLM distributed with Wrapyfi). I have also developed a framework called Wrapyfi for integrating robots and devices across multiple middleware including ROS, ROS 2, YARP, and ZeroMQ. To access all resources relating to Wrapyfi, visit https://modular.ml or the modular ML organization on Github.

Thanks for stopping by!


Jan 11, 2024 Our human participant study based on the Wrapyfi [code][paper] framework was accepted at the HRI ‘24 (Boulder, CO, USA) conference Late-Breaking Results (LBR)
Dec 16, 2023 Wrapyfi [code][paper] is now available on PyPi https://pypi.org/project/wrapyfi and the documentation on readthedocs https://wrapyfi.readthedocs.io
Dec 1, 2023 The Wrapyfi [code][paper] code contribution was accepted at the HRI ‘24 (Boulder, CO, USA) conference and will appear as an SC alongside the full-paper proceedings in March 2024

selected publications

  1. gasp_demo.gif
    GASP: Gated Attention for Saliency Prediction
    Fares Abawi, Tom Weber, and Stefan Wermter
    In Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI-21), 2021
  2. wrapyfi_demo.gif
    Wrapyfi: A Python Wrapper for Integrating Robots, Sensors, and Applications across Multiple Middleware
    Fares Abawi, Philipp Allgeuer, Di Fu, and 1 more author
    In Proceedings of the ACM/IEEE International Conference on Human-Robot Interaction (HRI ’24), 2024
  3. usp_preview.png
    Unified Dynamic Scanpath Predictors Outperform Individually Trained Neural Models
    Fares AbawiDi Fu, and Stefan Wermter
    CoRR, 2024