cs.ROOct 7, 2026

SiGNgapore - An Interactive Dataset for Sign-based Visual Navigation

Authors: Nicky Zimmerman, Joel Loo, Zishuo Wang, David Hsu

Organizations: National University of Singapore, SG

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

The future of autonomous robots in human-oriented environments depends on their ability to navigate in the absence of prebuilt maps; exploiting navigational aids, such as navigational signs, designed by humans for humans is critical for enabling autonomy. This manuscript introduces a unique dataset collected using a handheld device, at various public spaces across Singapore, representing environments that humans frequent daily, and focusing on sign-centric decision making for mapless navigation. The dataset includes RGB, sparse depth, odometry and IMU measurements of scenarios centered around navigational signs and the complex environments where they are placed. Additionally, we provide 56 long-horizon navigation missions and over 450 sign-centric scenarios. All data is provided in a human- readable format, as well as utility scripts for conversion to ROS 2. Lastly, we provide venue maps and GPS aligned scene graphs of the test environments. We discuss the potential use cases for this dataset.

Figures & tables

Figure 1 : Diversity of missions and environments in dataset. We collate a range of missions across indoor and outdoor environments, including common public spaces from malls to hospitals to parks. Indoor missions extend across multiple floors and interconnected buildings.
Figure 2 : Diversity of signs and scenes. Our dataset has diverse mission scenarios spanning a range of complex signs and scenes.
Figure 3 : Sign-optimal path. Given the mission goal ”COM1”, the robot must follow the sign-optimal path, which agrees with the navigational instruction associated with ”COM1” at every step of the navigation process. Other paths can also lead to the goal, but they are not considered sign-optimal, despite having equal (or lesser) step length.
Figure 4 : VQA format . The VQA options are provided in 3 different formats, to accommodate a variety of methods, from modular to end-to-end approaches: (a) as 3D points in the robot’s coordinate frame; (b) in pixel coordinates across multiple different view frames; (c) in a JSON format.
SensorData TypeData loggedFrequency
iPhone Camera1920 ×\times 1440 RGB images27474215 Hz\mathrm{Hz}
iPhone Lidar256 ×\times 192 depth images27474215 Hz\mathrm{Hz}
iPhone IMUTri-axial linear accelerations, tri-axial angular velocities1772208 sets100 Hz\mathrm{Hz}
iPhone Odometry(x, y, z, q x , q y , q z , q w )27474215 Hz\mathrm{Hz}
Table 1 : Raw data. Summary of the raw data collected in this dataset.
Figure 5 : GPS-aligned SceneGraphs. When available, we provide GPS-aligned SceneGrpah of the environment, annotated with building, floor and room nodes.
EnvironmentScenariosArea ( m2{\mathrm{m}}^{2} )TypeMissionsmulti-buildingmulti-floorOutdoorLength
E174-campusM1✗✓✓4
M2✓✓✗8
M3✓✓✗6
M4✓✓✗9
M5✓✓✗7
M6✓✓✗7
Table 2 : Environments and missions. For each environment, we provide information about its area size, type, number of grounding scenarios and the missions. Each mission can span across different floors, interconnected building and urban outdoor spaces. The mission length refers to number of scenarios in the sign-based optimal path.
File NameContentFormat
images/000000-999999.png1920 ×\times 1440 RGB imagePNG
depth/000000-999999.png256 ×\times 192 depth imagePNG
confidence/000000-999999.png256 ×\times 192 depth confidence mapPNG
imu.csvAccelerometer and gyroscope measurementsTimestamp ( s\mathrm{s} ), (x, y, z) linear accelerations m2 s−1{\mathrm{m}}^{2}\text{\,}{\mathrm{s}}^{-1} , (x, y, z) angular velocities ( rad s−1\mathrm{rad}\text{\,}{\mathrm{s}}^{-1} )
odometry.csvLocal pose estimatesTimestamp ( s\mathrm{s} ), frame id, (x, y, z) linear translation m\mathrm{m} , orientation (q x , q y , q z , q w ) rad\mathrm{rad}
camera_matrix.csv3 ×\times 3 camera intrinsics(f x , 0, c x , 0, f y , c y , 0, 0, 1)
Table 3 : Scenarios. Human-readable and image data formats and naming conventions for each scenario.
EnvironmentSuccess Rate (strict)Success RateVQA AccuracyParsing Success
E12/132/1357/93155/166
E20/40/414/2685/103
E32/102/1052/87109/134
E44/64/632/3479/90
E51/71/721/438/8
E60/70/723/4863/73
Table 4 : Evaluation. The performance of SignScene on the dataset, on the tasks of sign-based mapless navigation, visual sign grounding and sign parsing.

Explore similar work

CardsList

  1. Universal Navigation Interface: Robot-Free Data for Wheeled Robot Navigation

    Sep 17, 2026Sarvesh Prajapati, Ananya Trivedi, Lorena Maria Genua +3Robot NavigationRobotic Data Acquisition

    Institute for Experiential Robotics, Northeastern University, Boston, MA. · Khoury College of Computer Science, Northeastern University, Seattle, WA, USA.

  2. Bi3: A Biplatform, Bicultural, Biperson Dataset for Social Robot Navigation

    May 7, 2026Andrew Stratton, Phani Teja Singamaneni, Pranav Goyal +2Robot NavigationRobot Systems

    Department of Robotics, University of Michigan, Ann Arbor, USA · LAAS-CNRS, University of Toulouse, Toulouse, FR · INRIA, University of Lorraine, Nancy, FR

  3. POINav: Benchmarking and Enhancing Final-Meters Arrival in Real-World Vision-Language Navigation

    May 27, 2026Ruiyan Gong, Meisheng Zhang, Yuxiang Zhao +12Vision-Language NavigationReal-World

    Amap CV Lab, Alibaba Group · Peking University