HuSc3D: Human Sculpture dataset for 3D object reconstruction

2025-06-09

AI Summary

Existing 3D reconstruction benchmarks rely heavily on idealized synthetic or controlled real-world data, failing to reflect key challenges in outdoor mobile capture—such as dynamic backgrounds, smartphone white-balance artifacts, and imbalanced viewpoint distributions. To address this gap, we introduce HuSc3D: a novel benchmark comprising multi-view smartphone-captured images of six high-fidelity white figurine sculptures, accompanied by high-precision ground-truth 3D scans. HuSc3D is the first dataset to systematically integrate three realistic degradation factors—dynamic background interference, chromatic distortion, and geometrically minimal-texture surfaces—while enforcing highly imbalanced viewpoint sampling to stress-test model robustness under few-shot conditions, fine-structure recovery, and color ambiguity. Experiments demonstrate that HuSc3D effectively discriminates among state-of-the-art methods (e.g., NeRF and MVS), exposing systematic failures in hollow-structure reconstruction and few-shot generalization. It thus establishes a more diagnostic and rigorous benchmark for real-world 3D reconstruction.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

3D scene reconstruction from 2D images is one of the most important tasks in computer graphics. Unfortunately, existing datasets and benchmarks concentrate on idealized synthetic or meticulously captured realistic data. Such benchmarks fail to convey the inherent complexities encountered in newly acquired real-world scenes. In such scenes especially those acquired outside, the background is often dynamic, and by popular usage of cell phone cameras, there might be discrepancies in, e.g., white balance. To address this gap, we present HuSc3D, a novel dataset specifically designed for rigorous benchmarking of 3D reconstruction models under realistic acquisition challenges. Our dataset uniquely features six highly detailed, fully white sculptures characterized by intricate perforations and minimal textural and color variation. Furthermore, the number of images per scene varies significantly, introducing the additional challenge of limited training data for some instances alongside scenes with a standard number of views. By evaluating popular 3D reconstruction methods on this diverse dataset, we demonstrate the distinctiveness of HuSc3D in effectively differentiating model performance, particularly highlighting the sensitivity of methods to fine geometric details, color ambiguity, and varying data availability--limitations often masked by more conventional datasets.

Problem

• Addressing lack of realistic datasets for 3D reconstruction challenges
• Evaluating model sensitivity to geometric details and color ambiguity
• Benchmarking performance under varying image quantities and dynamic backgrounds

Innovation

• HuSc3D dataset for realistic 3D reconstruction challenges
• Features white sculptures with intricate perforations
• Varies image counts to test data sensitivity