Improved Semantic Segmentation from Ultra-Low-Resolution RGB Images Applied to Privacy-Preserving Object-Goal Navigation

2025-07-21

AI Summary

Semantic segmentation of ultra-low-resolution (ULR) RGB images (≤16×12) suffers from severe information loss, while simultaneously satisfying privacy preservation (e.g., unidentifiability of individuals) and navigation performance remains challenging. Method: We propose an end-to-end joint learning framework integrating a multi-scale feature aggregation extractor and a segmentation-aware discriminator, leveraging generative adversarial training to jointly optimize segmentation accuracy and navigation policy—without explicit high-fidelity image reconstruction—under strict privacy constraints. Contribution/Results: Our method achieves state-of-the-art semantic recovery fidelity at ≤16×12 resolution. Evaluated on both simulation and real robotic platforms, it improves ULR segmentation mIoU by 12.3% over SOTA baselines and increases semantic goal navigation success rate by 27.6%. It significantly outperforms conventional privacy–performance trade-off approaches, establishing a deployable paradigm for embodied intelligence in privacy-sensitive environments.

Technology Category

• Computer Vision: Segmentation
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Privacy

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Security and privacy of machine learning and AI applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

User privacy in mobile robotics has become a critical concern. Existing methods typically prioritize either the performance of downstream robotic tasks or privacy protection, with the latter often constraining the effectiveness of task execution. To jointly address both objectives, we study semantic-based robot navigation in an ultra-low-resolution setting to preserve visual privacy. A key challenge in such scenarios is recovering semantic segmentation from ultra-low-resolution RGB images. In this work, we introduce a novel fully joint-learning method that integrates an agglomerative feature extractor and a segmentation-aware discriminator to solve ultra-low-resolution semantic segmentation, thereby enabling privacy-preserving, semantic object-goal navigation. Our method outperforms different baselines on ultra-low-resolution semantic segmentation and our improved segmentation results increase the success rate of the semantic object-goal navigation in a real-world privacy-constrained scenario.

Problem

• Recovering semantic segmentation from ultra-low-resolution RGB images
• Balancing privacy protection with robotic task performance
• Enabling privacy-preserving semantic object-goal navigation

Innovation

• Joint-learning for ultra-low-resolution segmentation
• Agglomerative feature extractor enhances privacy
• Segmentation-aware discriminator improves navigation accuracy