Privacy-Preserving Semantic Segmentation from Ultra-Low-Resolution RGB Inputs

TL;DR

This paper presents a novel joint-learning framework for privacy-preserving semantic segmentation using ultra-low-resolution RGB inputs, balancing privacy and performance in real-world robotic tasks.

Contribution

It introduces a fully joint-learning approach that effectively mitigates visual degradation challenges in ultra-low-resolution semantic segmentation.

Findings

Our method outperforms baseline models in segmentation accuracy.

Ultra-low-resolution RGB inputs provide a good privacy-performance trade-off.

Successful deployment in robotic navigation demonstrates practical viability.

Abstract

RGB-based semantic segmentation has become a mainstream approach for visual perception and is widely applied in a variety of downstream tasks. However, existing methods typically rely on high-resolution RGB inputs, which may expose sensitive visual content in privacy-critical environments. Ultra-low-resolution RGB sensing suppresses sensitive information directly during image acquisition, making it an attractive privacy-preserving alternative. Nevertheless, recovering semantic segmentation from ultra-low-resolution RGB inputs remains highly challenging due to severe visual degradation. In this work, we introduce a novel fully joint-learning framework to mitigate the optimization conflicts exacerbated by visual degradation for ultra-low-resolution semantic segmentation. Experiments demonstrate that our method outperforms representative baselines in semantic segmentation performance and our ultra-low-resolution RGB input achieves a favorable trade-off between privacy preservation and semantic segmentation performance. We deploy our privacy-preserving semantic segmentation method in a real-world robotic object-goal navigation task, demonstrating successful downstream task execution even under severe visual degradation.