Multi-Domain Indoor Dataset for Visual Place Recognition and Anomaly Detection by Mobile Robots

Piotr Wozniak ¹, Tomasz Krzeszowski ², Bogdan Kwolek ³

⁰Faculty of Electrical and Computer Engineering, Rzeszow University of Technology, Rzeszow, 35-959, Poland. p.wozniak@prz.edu.pl.

Scientific Data +

|

May 19, 2025

View abstract on PubMed

Summary

A new dataset aids mobile robots in visual place recognition and anomaly detection. This resource accelerates research for autonomous systems navigating diverse indoor environments.

Area Of Science

Robotics
Computer Vision
Artificial Intelligence

Background

Visual location recognition, including place recognition (PR) and anomaly detection (AD), is essential for autonomous robots to determine their location and identify occupied spaces.
Existing datasets may not fully capture the complexities of real-world indoor environments for mobile robot navigation.

Purpose Of The Study

To introduce a comprehensive, multi-domain dataset for indoor visual place recognition and anomaly detection tailored for mobile robots.
To facilitate advancements in autonomous robot localization and environmental awareness.

Main Methods

Collected 89,550 RGB images across nine rooms, incorporating manual and robot-driven recordings.
Included diverse scenarios with variations in lighting, robot vision perspectives, and human activity.
Performed an analysis of existing literature datasets for comparative context.

Main Results

Achieved 80.18% accuracy in single-image anomaly detection using baseline methods.
Demonstrated 80.63%-84.18% accuracy for anomaly detection on image sequences.
Presented a detailed analysis of image sequence characteristics and key research findings.

Conclusions

The introduced dataset is a valuable, freely available resource for PR and AD research in mobile robotics.
Baseline methods show promising performance, highlighting the dataset's utility for evaluating new algorithms.
The dataset supports research into robust robot navigation and situational awareness in dynamic indoor settings.