SHREC 2025: Retrieval of Optimal Objects for Multi-modal Enhanced Language and Spatial Assistance (ROOMELSA)

Trong-Thuan Nguyen, Viet-Tham Huynh, Quang-Thuc Nguyen, Hoang-Phuc Nguyen, Long Bao Le, Thai Hoang Minh, Minh Nguyen Anh, Thang Nguyen Tien, Phat Nguyen Thuan, Huy Nguyen Phong, Bao Huynh Thai, Vinh-Tiep Nguyen, Duc-Vu Nguyen, Phu-Hoa Pham, Minh-Huy Le-Hoang, Nguyen-Khang Le, Minh-Chinh Nguyen, Minh-Quan Ho, Ngoc-Long Tran, Hien-Long Le-Hoang et al. (13 additional authors not shown)

Abstract: Highlights•In the ROOMELSA challenge, we formulate 3D object retrieval as a mask-conditioned, language-driven grounding task in 3D environments, unifying spatial reference and natural language understanding within a unified task.•We introduce ROOMELSA, a novel benchmark dataset comprising 1,622 scenes, 5,197 rooms, and 44,445 (mask, text) query pairs, with a dedicated unseen-category test split designed to evaluate cross-domain generalization.•We analyze the top-five challenge entries, revealing how scene-level inference, depth-aware reconstruction, and adaptive cross-modal fusion affect Mean Reciprocal Rank.