TRICKY 2025 HouseCat6D Object Pose Estimation Challenge with Specular and Transparent Surfaces

Weihang Li, Junwen Huang, HyunJun Jung, Guangyao Zhai, Pierluigi Zama Ramirez, Alex Costanzino, Fabio Tosi, Matteo Poggi, Luigi Di Stefano, Jean-Baptiste Weibel, Doris Antensteiner, Markus Vincze, Jing He, Yiqing Wang, Kexin Zhang, Licheng Jiao, Lingling Li, Fang Liu, Wenping Ma, Benjamin Busam; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2025, pp. 3292-3302

This paper presents the TRICKY 2025 HouseCat6D Category-Level Object Pose Estimation Challenge, held in conjunction with the ICCV 2025 workshop on Transparent and Reflective Objects in the Wild. The challenge addresses the critical limitations of existing pose estimation systems when applied to non-Lambertian surfaces, such as glass and metal. Leveraging the HouseCat6D dataset comprising realistic home environments with a diverse range of transparent and specular objects, the challenge pushes state-of-the-art algorithms to estimate object pose, scale, and shape in photometrically complex scenes. Unlike traditional benchmarks focused on texture-rich, opaque objects, this challenge emphasizes robustness under reflective highlights, refractions, and partial transparency. By promoting research in these underexplored conditions, the challenge contributes toward generalizable and category-level object understanding in unconstrained real-world settings.