๐ค AI Summary
This study addresses the challenge of balancing privacy preservation and estimation accuracy in camera-based HVAC thermal comfort control. To this end, it proposes a Vision Transformerโbased method for clothing classification directly on encrypted images, marking the first application of Vision Transformers to the task of estimating clothing insulation values in the encrypted domain. By operating on encrypted visual data, the approach safeguards user visual privacy while circumventing the significant performance degradation typically caused by conventional pixel-level privacy-preserving techniques. Experimental results on the DeepFashion dataset demonstrate that the proposed method achieves classification accuracy comparable to that obtained with original, unencrypted images across all insulation categories, thereby unifying lossless privacy protection with high-precision thermal estimation.
๐ 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.