Vision-centric Robotic Perception for Real-world Productivity

Please, join the ME Department on Monday, June 16th for a presentation from Dr. Huaijin Chen, Assistant Professor of Computer Science at the University of Hawai?i at Manoa.


How can visual sensing drive productivity in real-world environments such as farms and factory floors?
This talk presents recent work from my lab on vision-centric robotic perception, with a focus on integrating advanced imaging modalities into field-deployable systems. I will begin with the recent Perception for Industrial Robotics Automation (PIRA) Workshop at CVPR 2025, which we co-organized with OpenCV and Intrinsic (an Alphabet company) to explore RGB, infrared, and polarization imaging for robotic bin picking. The workshop highlights key challenges in multimodal sensor fusion, low-latency inference, and closed-loop perception for industrial automation. Next, I describe our winning project in the 2025 Farm Robotics Challenge: a low-cost, land + aerial camera-based farm intelligence system that delivers coarse-to-fine farm information, from field-level flower counts to individual fruit sizing. Finally, I introduce PolarTact3D, a multimodal perception system that leverages polarization imaging for tactile sensing to enable dense color 3D surface reconstruction, an essential capability for future field robots operating in complex physical environments.