Investigating Domain Gaps for Indoor 3D Object Detection

2025-08-24

AI Summary

This paper addresses domain shift in indoor 3D object detection—arising from synthetic-vs-real data discrepancies, point cloud quality variations, and inconsistencies in scene layout and instance distributions. We systematically investigate multiple domain gaps and their impacts. To this end, we introduce two large-scale synthetic datasets, ProcTHOR-OD and ProcFront, and establish a unified cross-dataset benchmark covering ScanNet, SUN RGB-D, and 3D-Front. We present the first decoupled analysis and adaptation study for three key domain factors: point cloud fidelity, scene layout, and instance distribution—specifically tailored to indoor 3D detection. Furthermore, we integrate diverse domain adaptation methods and provide reproducible baseline results. Experiments demonstrate that our benchmark and analytical framework significantly improve cross-domain generalization of 3D detectors, laying a foundation for future research in domain adaptation for 3D object detection.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Internet of Things, Sensor Networks & Smart Cities

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

As a fundamental task for indoor scene understanding, 3D object detection has been extensively studied, and the accuracy on indoor point cloud data has been substantially improved. However, existing researches have been conducted on limited datasets, where the training and testing sets share the same distribution. In this paper, we consider the task of adapting indoor 3D object detectors from one dataset to another, presenting a comprehensive benchmark with ScanNet, SUN RGB-D and 3D Front datasets, as well as our newly proposed large-scale datasets ProcTHOR-OD and ProcFront generated by a 3D simulator. Since indoor point cloud datasets are collected and constructed in different ways, the object detectors are likely to overfit to specific factors within each dataset, such as point cloud quality, bounding box layout and instance features. We conduct experiments across datasets on different adaptation scenarios including synthetic-to-real adaptation, point cloud quality adaptation, layout adaptation and instance feature adaptation, analyzing the impact of different domain gaps on 3D object detectors. We also introduce several approaches to improve adaptation performances, providing baselines for domain adaptive indoor 3D object detection, hoping that future works may propose detectors with stronger generalization ability across domains. Our project homepage can be found in https://jeremyzhao1998.github.io/DAVoteNet-release/.

Problem

• Investigating domain gaps in indoor 3D object detection across datasets
• Adapting detectors from synthetic to real indoor point cloud data
• Analyzing impact of domain gaps like quality and layout variations

Innovation

• Domain adaptive 3D object detection across datasets
• Synthetic-to-real adaptation using simulator-generated datasets
• Analyzing domain gaps impact on detector generalization