SAMURAI: Shape-Aware Multimodal Retrieval for 3D Object Identification

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Vo, Dinh-Khoi, Nguyen, Van-Loc, Tran, Minh-Triet, Le, Trung-Nghia
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908423338065920
author Vo, Dinh-Khoi
Nguyen, Van-Loc
Tran, Minh-Triet
Le, Trung-Nghia
author_facet Vo, Dinh-Khoi
Nguyen, Van-Loc
Tran, Minh-Triet
Le, Trung-Nghia
contents Retrieving 3D objects in complex indoor environments using only a masked 2D image and a natural language description presents significant challenges. The ROOMELSA challenge limits access to full 3D scene context, complicating reasoning about object appearance, geometry, and semantics. These challenges are intensified by distorted viewpoints, textureless masked regions, ambiguous language prompts, and noisy segmentation masks. To address this, we propose SAMURAI: Shape-Aware Multimodal Retrieval for 3D Object Identification. SAMURAI integrates CLIP-based semantic matching with shape-guided re-ranking derived from binary silhouettes of masked regions, alongside a robust majority voting strategy. A dedicated preprocessing pipeline enhances mask quality by extracting the largest connected component and removing background noise. Our hybrid retrieval framework leverages both language and shape cues, achieving competitive performance on the ROOMELSA private test set. These results highlight the importance of combining shape priors with language understanding for robust open-world 3D object retrieval.
format Preprint
id arxiv_https___arxiv_org_abs_2506_21056
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SAMURAI: Shape-Aware Multimodal Retrieval for 3D Object Identification
Vo, Dinh-Khoi
Nguyen, Van-Loc
Tran, Minh-Triet
Le, Trung-Nghia
Computer Vision and Pattern Recognition
Retrieving 3D objects in complex indoor environments using only a masked 2D image and a natural language description presents significant challenges. The ROOMELSA challenge limits access to full 3D scene context, complicating reasoning about object appearance, geometry, and semantics. These challenges are intensified by distorted viewpoints, textureless masked regions, ambiguous language prompts, and noisy segmentation masks. To address this, we propose SAMURAI: Shape-Aware Multimodal Retrieval for 3D Object Identification. SAMURAI integrates CLIP-based semantic matching with shape-guided re-ranking derived from binary silhouettes of masked regions, alongside a robust majority voting strategy. A dedicated preprocessing pipeline enhances mask quality by extracting the largest connected component and removing background noise. Our hybrid retrieval framework leverages both language and shape cues, achieving competitive performance on the ROOMELSA private test set. These results highlight the importance of combining shape priors with language understanding for robust open-world 3D object retrieval.
title SAMURAI: Shape-Aware Multimodal Retrieval for 3D Object Identification
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.21056