摘要
This paper presents the submission of our RU-DS team to the Pixel Privacy Task 2020. We propose to fool the blind image quality assessment model by transforming images based on optimizing a human-understandable color filter. In contrast to the common work that relies on small, Lp-bounded additive pixel perturbations, our approach yields large yet smooth perturbations. Experimental results demonstrate that in the specific context of this task, our approach is able to achieve strong adversarial effects, but has to sacrifice the image appeal.
| 源语言 | 英语 |
|---|---|
| 期刊 | CEUR Workshop Proceedings |
| 卷 | 2882 |
| 出版状态 | 已出版 - 2020 |
| 已对外发布 | 是 |
| 活动 | Multimedia Evaluation Benchmark Workshop 2020, MediaEval 2020 - Virtual, Online 期限: 14 12月 2020 → 15 12月 2020 |
学术指纹
探究 'Fooling blind image quality assessment by optimizing a human-understandable color filter' 的科研主题。它们共同构成独一无二的学术指纹。引用此
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver