Privacy-Preserving Clothing Classification using Vision Transformer for Thermal Comfort Estimation

arXiv cs.CV / 4/30/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes a privacy-preserving clothing classification approach aimed at enabling secure occupant-centric control (OCC) for HVAC systems.
It notes that prior HVAC comfort-control research used camera images but largely ignored privacy protection for occupant imagery, leaving a key gap.
The method uses a Vision Transformer (ViT) tailored to clothing insulation estimation and is designed to work effectively on encrypted images.
Experiments on the DeepFashion dataset show that conventional pixel-based privacy techniques cause a major accuracy drop, whereas the proposed scheme preserves high accuracy across all clothing-insulation categories.
Overall, the study suggests Vision-Transformer-based privacy-preserving inference can avoid the typical accuracy degradation seen in earlier privacy-preserving image classification methods.

Abstract

A privacy-preserving clothing classification scheme is presented to enable secure occupant-centric control (OCC) systems. Although the utilization of camera images for HVAC control has been widely studied to optimize thermal comfort, privacy protection of occupant images has not been considered in prior works. While various privacy-preserving methods have been proposed for image classification, applying conventional schemes results in severe accuracy degradation. In this paper, we introduce a privacy-preserving classification method using Vision Transformer (ViT) applied to clothing insulation estimation. In an experiment using the DeepFashion dataset categorized by clothing insulation, while the conventional pixel-based method suffers a severe accuracy drop, our scheme maintains a high accuracy on encrypted images, showing no degradation from plain images across all categories.