Quantum gated hybrid model shrinks underwater sonar detection size
Quantum-Gated LiteSSD: A Parameter-Efficient Lightweight Hybrid Quantum-Classical Framework for Forward-Looking Sonar Object Detection
Computer Vision and Pattern Recognition
Summary
Detecting objects underwater using forward-looking sonar is important but hard to do on small devices that need compact software. The authors created Quantum-Gated LiteSSD, a hybrid approach that combines quantum computing ideas with classical methods to make a detector that uses far fewer parameters while keeping good accuracy. They tested it on underwater sonar datasets and found it achieves high accuracy with much smaller model sizes compared to popular existing detectors. This makes it easier to deploy on small embedded platforms for underwater sensing.
What this means in practice
- •For embedded system developers: Build smaller sonar object detectors that fit and run efficiently on hardware with limited memory and power.
- •For robotics engineers: Integrate compact, accurate forward-looking sonar models in underwater robots for debris and obstacle detection.
Authors
Niloy Kumar Mondal, Poulomi Sarker Puja
Abstract
Forward-looking sonar object detection is essential for underwater perception, yet deployment on embedded platforms requires highly compact models. To address this challenge, we explore quantum computing and introduce Quantum-Gated LiteSSD, a parameter-efficient hybrid quantum--classical detector that reformulates QuCNet-style multi-circuit quantum processing as an identity-centered channel-gating mechanism for spatial feature modulation. Experiments on the Marine Debris Watertank dataset and UATD forward-looking sonar benchmarks demonstrate an effective parameter--accuracy trade-off. The proposed detector achieves 90.84% $\mathrm{mAP}_{50}$ on Watertank with approximately $62\times$ fewer parameters than YOLO26s and $164.3\times$ fewer parameters than SSD-VGG16. On UATD, the model achieves 70.37% $\mathrm{mAP}_{50}$ with only 0.150M parameters, making it approximately $4.1\times$ smaller than SSGA-YOLO while retaining meaningful multi-class detection capability.