🏆 Best Paper Award (3rd) — RSS 2025 Workshop on Navigating Contact Dynamics in Robotics

PolarTact3D: Single-shot Tactile 3D Shape and Color Sensing with Polarization Imaging

Kai Garcia, Mairi Yoshioka, Huaijin Chen, Tyler Ray, Tianlu Wang, Frances Zhu
University of Hawaii at Mānoa • Colorado School of Mines
RSS 2025 | Los Angeles, USA
Paper Code
System Pipeline Diagram

System Overview. Our setup uses a polarization camera mounted in an aluminum cage with side illumination. A glossy, transparent polyethylene (PE) film is stretched across the cage bottom. When pressed onto a surface, the film conforms to the object, acting as a dielectric layer that enables shape-from-polarization (SfP) reconstruction via either neural networks or physics-based methods.

Abstract

Vision-based tactile sensors are essential for robotic applications such as grasping and physical interaction. We propose a low-cost, polarization-imaging-based tactile sensor that captures both shape and color information in a single shot.

Unlike photometric-stereo-based solutions like GelSight, which require precise internal illumination and cannot reliably capture color due to the reflective coating required for surface reconstruction, our method leverages the Angle of Linear Polarization (AoLP) and Degree of Linear Polarization (DoLP) to encode 3D geometry.

This approach enables robust shape reconstruction, even on transparent or specular targets. The sensor is constructed using commercial transparent polyethylene (PE) film and an off-the-shelf polarization camera, making it simple and inexpensive to build.

We demonstrate the effectiveness of our design through real-world experiments on various contact surface scenarios.

Poster

Poster thumbnail

Tap to view full poster (PDF)

BibTeX


        @inproceedings{garcia2025polartact3d,
            author    = {Kai Garcia and Mairi Yoshioka and Tyler Ray and Tianlu Wang and Frances Zhu and Huaijin Chen },
            title     = {{PolarTact3D}: Single-shot Tactile 3-D Shape and Color Sensing with Polarization Imaging},
            booktitle = {Workshop on Navigating Contact Dynamics in Robotics},
            series = {Robotics: Science and Systems (RSS)},
            year      = {2025},
            address   = {Los Angeles, CA},
            pages     = {1-1},
            note = {Best Paper Award (3rd Place)}
          }