Computer Science > Computer Vision and Pattern Recognition
[Submitted on 30 Apr 2024 (v1), last revised 15 Dec 2024 (this version, v6)]
Title:Espresso: Robust Concept Filtering in Text-to-Image Models
View PDF HTML (experimental)Abstract:Diffusion based text-to-image models are trained on large datasets scraped from the Internet, potentially containing unacceptable concepts (e.g., copyright-infringing or unsafe). We need concept removal techniques (CRTs) which are i) effective in preventing the generation of images with unacceptable concepts, ii) utility-preserving on acceptable concepts, and, iii) robust against evasion with adversarial prompts. No prior CRT satisfies all these requirements simultaneously. We introduce Espresso, the first robust concept filter based on Contrastive Language-Image Pre-Training (CLIP). We identify unacceptable concepts by using the distance between the embedding of a generated image to the text embeddings of both unacceptable and acceptable concepts. This lets us fine-tune for robustness by separating the text embeddings of unacceptable and acceptable concepts while preserving utility. We present a pipeline to evaluate various CRTs to show that Espresso is more effective and robust than prior CRTs, while retaining utility.
Submission history
From: Anudeep Das [view email][v1] Tue, 30 Apr 2024 03:13:06 UTC (1,590 KB)
[v2] Wed, 1 May 2024 18:30:14 UTC (1,669 KB)
[v3] Wed, 8 May 2024 00:22:32 UTC (3,330 KB)
[v4] Fri, 7 Jun 2024 14:28:24 UTC (2,163 KB)
[v5] Mon, 9 Sep 2024 16:51:21 UTC (2,175 KB)
[v6] Sun, 15 Dec 2024 16:20:37 UTC (10,094 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.