Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FairAdapter: Detecting AI-generated Images with Improved Fairness

Feng Ding, Jun Zhang, Xinan He, Jianfeng Xu

TL;DR

A novel framework named Fairadapter is proposed to tackle the issue of detection fairness, and in comparison with existing state-of-the-art methods, this model achieves improved fairness performance.

Abstract

The high-quality, realistic images generated by generative models pose significant challenges for exposing them.So far, data-driven deep neural networks have been justified as the most efficient forensics tools for the challenges. However, they may be over-fitted to certain semantics, resulting in considerable inconsistency in detection performance across different contents of generated samples. It could be regarded as an issue of detection fairness. In this paper, we propose a novel framework named Fairadapter to tackle the issue. In comparison with existing state-of-the-art methods, our model achieves improved fairness performance. Our project: https://github.com/AppleDogDog/FairnessDetection

FairAdapter: Detecting AI-generated Images with Improved Fairness

TL;DR

A novel framework named Fairadapter is proposed to tackle the issue of detection fairness, and in comparison with existing state-of-the-art methods, this model achieves improved fairness performance.

Abstract

The high-quality, realistic images generated by generative models pose significant challenges for exposing them.So far, data-driven deep neural networks have been justified as the most efficient forensics tools for the challenges. However, they may be over-fitted to certain semantics, resulting in considerable inconsistency in detection performance across different contents of generated samples. It could be regarded as an issue of detection fairness. In this paper, we propose a novel framework named Fairadapter to tackle the issue. In comparison with existing state-of-the-art methods, our model achieves improved fairness performance. Our project: https://github.com/AppleDogDog/FairnessDetection

Paper Structure

This paper contains 16 sections, 7 equations, 1 figure, 4 tables.

Table of Contents

  1. Introduction
  2. RELATED WORKS
  3. Image generation models
  4. Forensics for AI-generated images
  5. METHODOLOGY
  6. Motivation
  7. Overview of Proposed Method
  8. FairAdapter Network
  9. ClassifyAdapter network
  10. Categories Loss Function
  11. FairAdapter Loss
  12. ClassifyAdapter Loss
  13. EXPERIMENTS
  14. Experimental Settings
  15. Experimental Results
  16. ...and 1 more sections

Figures (1)

  • Figure 1: An overview of our proposed method: 1) For the input stage, we select one category of natural and AI-generated images, along with natural images of other categories, to input into the image encoder. 2) In the FairnessAssistant module, we compute the FairAdapter loss by mixing the image semantics. 3) In the Classification module, we feed the enhanced semantics into the ClassifyAdapter network to complete the classification task.