ICRA 2026 Accpeted

ScaleMaster Dataset

A Large-Scale Dataset and Benchmark for Evaluating Scale Consistency in Complex Indoor Environments

Hyoseok Ju1, Bokeon Suh1, Giseop Kim1†
1Daegu Gyeongbuk Institute of Science and Technology, Daegu, Republic of Korea (DGIST)
💻 GitHub 📥 Download Dataset

Abstract

Recent advances in deep monocular visual SLAM have achieved impressive accuracy and dense reconstruction capabilities, yet their robustness to scale inconsistency in large- scale indoor environments remains largely unexplored. Existing benchmarks are limited to room-scale or structurally simple settings, leaving critical issues of intra-session scale drift and inter-session scale ambiguity insufficiently addressed. To fill this gap, we introduce the ScaleMaster Dataset, the first benchmark explicitly designed to evaluate scale consistency under challenging scenarios such as multi-floor structures, long trajectories, repetitive views, and low-texture regions. We systematically analyze the vulnerability of state-of-the-art deep monocular visual SLAM systems to scale inconsistency, providing both qualitative and quantitative evaluations. Crucially, our analysis extends beyond traditional trajectory metrics to include a direct map-to-map quality assessment using metrics like Chamfer distance against high-fidelity 3D ground truth. Our results reveal that while these traditional methods demonstrate strong performance on existing benchmarks, they suffer from severe scale-related failures in realistic, large-scale indoor environments. By releasing the ScaleMaster dataset and baseline results, we aim to establish a foundation for future research toward developing scale-consistent and reliable visual SLAM systems.

Dataset Statistics

25 Sequences
3.8 km+ Total Length
10+ Environments
RGB+D+IMU Data Type

Dataset Sequences

Click on any sequence to explore in 3D

📚 Library Environments (9 sequences)

🏢 Large Hall Environments (5 sequences)

🚗 Parking & Basement (3 sequences)

🪜 Stairs & Station (3 sequences)

🏠 Indoor Rooms (5 sequences)

Download Dataset

📥 Download

Survey for Download
Individual sequences
Browse Dataset

💻 Git Clone

Clone entire dataset
with Git LFS
Show Command

📖 Documentation

Dataset format
Usage examples
View on GitHub

Citation

@inproceedings{ju2026scalemaster,
  title={Have We Mastered Scale in Deep Monocular Visual SLAM? The ScaleMaster Dataset and Benchmark},
  author={Ju, Hyoseok and Suh, Bokeon and Kim, Giseop},
  booktitle={Proceedings of the IEEE International Conference on Robotics and Automation (ICRA)},
  year={2026}
}
Loading Point Cloud...
0%