Commit
•
983da52
1
Parent(s):
4460b58
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,44 +1,58 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
| [
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
---
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
tags:
|
| 6 |
+
- art
|
| 7 |
+
- unlearned_textencoder
|
| 8 |
+
- safe_diffusion
|
| 9 |
+
- adversarial_training
|
| 10 |
+
- defensive_unlearning
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
# Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models
|
| 15 |
+
### [Project Website]() | [Arxiv Preprint](https://arxiv.org/abs/2405.15234) | [Fine-tuned Weights](https://drive.google.com/drive/folders/1Nf-EJ2W3CsZwpc5blZFi7tm7o1wEiTg4?usp=sharing) | [Demo]() <br>
|
| 16 |
+
Our proposed robust unlearning framework, AdvUnlearn, enhances diffusion models' safety by robustly erasing unwanted concepts through adversarial training, achieving an optimal balance between concept erasure and image generation quality.
|
| 17 |
+
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
<div align="center">
|
| 21 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/63e44e62789dcaae43c865d9/vad9l9ME0KD0OJKYLcFun.png" />
|
| 22 |
+
</div>
|
| 23 |
+
|
| 24 |
+
|
| 25 |
+
## Baselines
|
| 26 |
+
|
| 27 |
+
| DM Unlearning Methods | Nudity | Van Gogh | Objects |
|
| 28 |
+
|:-------|:----:|:-------:| :-------:|
|
| 29 |
+
| [ESD](https://github.com/rohitgandikota/erasing) (Erased Stable Diffusion) | ✅ | ✅ | ✅
|
| 30 |
+
| [FMN](https://github.com/SHI-Labs/Forget-Me-Not) (Forget-Me-Not) | ✅ | ✅ | ✅
|
| 31 |
+
| [AC](https://github.com/nupurkmr9/concept-ablation) (Ablating Concepts) | ❌ | ✅ | ❌
|
| 32 |
+
| [UCE](https://github.com/rohitgandikota/unified-concept-editing) (Unified Concept Editing) | ✅ | ✅ | ❌
|
| 33 |
+
| [SalUn](https://github.com/OPTML-Group/Unlearn-Saliency) (Saliency Unlearning) | ✅ | ❌ | ✅
|
| 34 |
+
| [SH](https://github.com/JingWu321/Scissorhands_ex) (ScissorHands) | ✅ | ❌ | ✅
|
| 35 |
+
| [ED](https://github.com/JingWu321/EraseDiff) (EraseDiff) | ✅ | ❌ | ✅
|
| 36 |
+
| [SPM](https://github.com/Con6924/SPM) (concept-SemiPermeable Membrane) | ✅ | ✅ | ✅
|
| 37 |
+
| **AdvUnlearn (Ours)** | ✅ | ✅ | ✅
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
<br>
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
## Cite Our Work
|
| 44 |
+
The preprint can be cited as follows:
|
| 45 |
+
```
|
| 46 |
+
@misc{zhang2024defensive,
|
| 47 |
+
title={Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models},
|
| 48 |
+
author={Yimeng Zhang and Xin Chen and Jinghan Jia and Yihua Zhang and Chongyu Fan and Jiancheng Liu and Mingyi Hong and Ke Ding and Sijia Liu},
|
| 49 |
+
year={2024},
|
| 50 |
+
eprint={2405.15234},
|
| 51 |
+
archivePrefix={arXiv},
|
| 52 |
+
primaryClass={cs.CV}
|
| 53 |
+
}
|
| 54 |
+
```
|
| 55 |
+
|
| 56 |
+
---
|
| 57 |
+
license: cc-by-4.0
|
| 58 |
---
|