Beyond Pixels: Enhancing LIME with Hierarchical Features and Segmentation Foundation Models

2024-03-12

AI Summary

To address semantic distortion and poor interpretability in LIME-based image explanations caused by coarse-grained segmentation, this paper proposes a hierarchical segmentation enhancement framework that integrates foundation models with data-driven techniques. Methodologically: (1) segmentation foundation models (e.g., SAM) are leveraged to generate high-fidelity semantic regions; (2) a data-driven, hierarchical graph-cut–superpixel collaborative segmentation mechanism is designed, enabling user-controllable explanation granularity via interactive refinement; (3) LIME’s local linear approximation is embedded to preserve explanatory consistency. The core contribution is the first introduction of a semantics-aligned hierarchical segmentation paradigm, substantially improving alignment between explanations and human cognition as well as ground-truth semantic concepts. Our approach outperforms mainstream XAI baselines across multiple quantitative evaluation metrics—including faithfulness, localization accuracy, and human agreement—and the implementation is publicly available.

Technology Category

• Computer Vision: Segmentation
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

LIME (Local Interpretable Model-agnostic Explanations) is a popular XAI framework for unraveling decision-making processes in vision machine-learning models. The technique utilizes image segmentation methods to identify fixed regions for calculating feature importance scores as explanations. Therefore, poor segmentation can weaken the explanation and reduce the importance of segments, ultimately affecting the overall clarity of interpretation. To address these challenges, we introduce the DSEG-LIME (Data-Driven Segmentation LIME) framework, featuring: i) a data-driven segmentation for human-recognized feature generation by foundation model integration, and ii) a user-steered granularity in the hierarchical segmentation procedure through composition. Our findings demonstrate that DSEG outperforms on several XAI metrics on pre-trained ImageNet models and improves the alignment of explanations with human-recognized concepts. The code is available under: https://github. com/patrick-knab/DSEG-LIME

Problem

• LIME
• Image Recognition
• Interpretability

Innovation

• DSEG-LIME
• Image Recognition
• Interpretable AI