IROS 2025

The IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025) takes place on 19–25 October 2025 in Hangzhou, China.

At IROS 2025, Taufiq Ahmed from the University of Oulu will present a new method titled Exponentially Weighted Instance-Aware Repeat Factor Sampling (E-IRFS) — a lightweight and effective strategy to improve model training in long-tailed object detection, a key challenge in real-world UAV-based surveillance scenarios.

The method strengthens rare-class representation in training data without introducing instability and is especially impactful for compact models such as YOLOv11-Nano. Experiments show that E-IRFS improves detection of under-represented classes like smoke, water, or people — particularly valuable in emergency-response and environmental-monitoring missions.

This research supports 6G Flagship’s Distributed Intelligence by addressing key challenges including robust AI training under data imbalance, efficient edge deployment in multi-drone systems, and enhanced real-time situational awareness.

Visit the official IROS 2025 website for the full programme and updates. Code and dataset details are available on
GitHub.

View other events featuring 6G Flagship research on our event calendar.