Vision Transformer enables privacy-preserving clothing classification for thermal comfort

Researchers have developed a novel privacy-preserving method for classifying clothing types using Vision Transformers. This approach aims to enable secure occupant-centric control systems for optimizing thermal comfort without compromising user privacy. Experiments on the DeepFashion dataset demonstrated that the proposed scheme maintains high accuracy on encrypted images, unlike conventional methods that suffer significant degradation. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a privacy-preserving technique for vision models that could enhance the security of smart building systems.

RANK_REASON Academic paper introducing a new method for privacy-preserving image classification.

Read on arXiv cs.CV →

• Vision Transformer
• DeepFashion
• arXiv

• paper
• safety