Detecting Spurious Correlations via Robust Visual Concepts in Real and AI-Generated Image Classification

Published: 27 Oct 2023, Last Modified: 16 Nov 2023NeurIPS XAIA 2023EveryoneRevisionsBibTeX
TL;DR: An efficient method, compatible with both real and AI-generated images, for detecting spurious correlations in image classification.
Abstract: Often machine learning models tend to automatically learn associations present in the training data without questioning their validity or appropriateness. This undesirable property is the root cause of the manifestation of spurious correlations, which render models unreliable and prone to failure in the presence of distribution shifts. Research shows that most methods attempting to remedy spurious correlations are only effective for a model's known spurious associations. Current spurious correlation detection algorithms either rely on extensive human annotations or are too restrictive in their formulation. Moreover, they rely on strict definitions of visual artifacts that may not apply to data produced by generative models, as they are known to hallucinate contents that do not conform to standard specifications. In this work, we introduce a general-purpose method that efficiently detects potential spurious correlations, and requires significantly less human interference in comparison to the prior art. Additionally, the proposed method provides intuitive explanations while eliminating the need for pixel-level annotations. We demonstrate the proposed method's tolerance to the peculiarity of AI-generated images, which is a considerably challenging task, one where most of the existing methods fall short. Consequently, our method is also suitable for detecting spurious correlations that may propagate to downstream applications originating from generative models.
Submission Track: Full Paper Track
Application Domain: Computer Vision
Survey Question 1: Spurious correlations in machine learning models pose numerous challenges, including issues with generalization and robustness. There is a strong need for an effective technique capable of identifying spurious associations learned by image classifiers, suitable for both real and AI-generated images. This work leverages concepts from Explainable AI (XAI) to construct a robust spurious correlation detection method.
Survey Question 2: Building on explainable AI (XAI) concepts has been shown to be a viable approach for detecting spurious correlations. Apart from this, visualizing spurious patterns could be helpful for understanding and remedying spurious associations. In our work, use concepts from XAI to do both of these tasks.
Survey Question 3: We propose a new technique (named RF-CAM) that draws on concepts from SHAP and GradCAM, and we also use vanilla GradCAM for visualization purposes.
Submission Number: 14