[paper]SPATIALLY TRANSFORMED ADVERSARIAL EXAMPLES
本文提出了一種基於空間變換（spatially transformed）（smooth image deformation）的對抗樣本生成算法。（ L p L_p Lp距離較大，但圖像更真實更不容易被檢測） 之前的生成算法大多是將 L p L_p Lp範數作爲損失函數，然後直接調整像素值。 兩個重要因素 1）圖片亮度和質量 2）幾何關係 先前的大多數對抗攻擊都建立在改變亮度和質量的基礎上，同時
相關文章
- 1. SPATIALLY TRANSFORMED ADVERSARIAL EXAMPLES
- 2. Paper Review: Adversarial Examples
- 3. [advGAN]Generating Adversarial Examples With Adversarial Networks
- 4. Generating Adversarial Examples with Adversarial Networks
- 5. Explaining and Harnessing Adversarial Examples
- 6. CS231n Lecture 16 | Adversarial Examples and Adversarial Training
- 7. Generating Adversarial Examples with Adversarial Networks閱讀筆記
- 8. Adversarial Examples 對抗樣本
- 9. [paper]ADVERSARIAL EXAMPLES IN THE PHYSICAL WORLD
- 10. explaining and harnessing adversarial examples(FGSM)
- 更多相關文章...
- • XPath Examples - XPath 教程
- • Rust 註釋 - RUST 教程
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. SPATIALLY TRANSFORMED ADVERSARIAL EXAMPLES
- 2. Paper Review: Adversarial Examples
- 3. [advGAN]Generating Adversarial Examples With Adversarial Networks
- 4. Generating Adversarial Examples with Adversarial Networks
- 5. Explaining and Harnessing Adversarial Examples
- 6. CS231n Lecture 16 | Adversarial Examples and Adversarial Training
- 7. Generating Adversarial Examples with Adversarial Networks閱讀筆記
- 8. Adversarial Examples 對抗樣本
- 9. [paper]ADVERSARIAL EXAMPLES IN THE PHYSICAL WORLD
- 10. explaining and harnessing adversarial examples(FGSM)