DiffInf: Influence-Guided Diffusion for Supervision Alignment in Facial Attribute Learning

2026-03-06

AI Summary

This work addresses the challenge of label-image inconsistency in facial attribute classification, often caused by subjective annotations and visual distractions, which introduces erroneous supervisory signals. To mitigate this issue, the authors propose a self-influence-guided diffusion rectification framework that leverages a lightweight influence predictor to identify high-influence yet mislabeled samples. These samples are then refined through a latent diffusion autoencoder, which performs generative editing to align image content with their assigned labels while preserving identity characteristics and photorealism. Crucially, the approach avoids discarding any training samples, thereby maintaining the integrity of the original data distribution, and enables end-to-end optimization via differentiable influence regularization. Experimental results demonstrate that the method significantly outperforms existing robust training strategies for noisy labels, influence-based filtering techniques, and standard baselines across multiple facial attribute classification tasks, leading to improved model generalization.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Facial attribute classification relies on large-scale annotated datasets in which many traits, such as age and expression, are inherently ambiguous and continuous but are discretized into categorical labels. Annotation inconsistencies arise from subjectivity and visual confounders such as pose, illumination, expression, and demographic variation, creating mismatch between images and assigned labels. These inconsistencies introduce supervision errors that impair representation learning and degrade downstream prediction. We introduce DiffInf, a self-influence--guided diffusion framework for mitigating annotation inconsistencies in facial attribute learning. We first train a baseline classifier and compute sample-wise self-influence scores using a practical first-order approximation to identify training instances that disproportionately destabilize optimization. Instead of discarding these influential samples, we apply targeted generative correction via a latent diffusion autoencoder to better align visual content with assigned labels while preserving identity and realism. To enable differentiable guidance during correction, we train a lightweight predictor of high-influence membership and use it as a surrogate influence regularizer. The edited samples replace the originals, yielding an influence-refined dataset of unchanged size. Across multi-class facial attribute classification, DiffInf consistently improves generalization compared with standard noisy-label training, robust optimization baselines, and influence-based filtering. Our results demonstrate that repairing influential annotation inconsistencies at the image level enhances downstream facial attribute classification without sacrificing distributional coverage.

Problem

• annotation inconsistency
• facial attribute learning
• supervision error
• label noise
• visual confounders

Innovation

• influence-guided diffusion
• annotation inconsistency
• facial attribute learning
• latent diffusion autoencoder
• self-influence scoring