Motivation

The rise of deep learning has ushered in significant progress in computer vision (CV) tasks, yet the “black box” nature of these models often precludes interpretability. This challenge has spurred the development of Explainable Artificial Intelligence (XAI) that generates explanations to AI’s decision-making process. An explanation is aimed to not only faithfully reflect the true reasoning process (i.e., faithfulness) but also align with humans’ reasoning (i.e., alignment). Within XAI, particularly in image processing, visual explanations highlights images’ critical areas important to predictions to elucidate the reasoning behind machine learning models. Despite the considerable body of research in visual explanations, standardized benchmarks for evaluating them are seriously underdeveloped. To address this issue, we develop a benchmark for visual explanation, consisting of eight datasets with human explanation annotations from various domains, accommodating both post-hoc and intrinsic visual explanation methods. Additionally, we devise a visual explanation pipeline that includes data loading, explanation generation, and method evaluation. Our proposed benchmarks facilitate a fair evaluation and comparison of visual explanation methods. Building on our curated collection of datasets, we benchmarked eight existing visual explanation methods and conducted a thorough comparison across four selected datasets using six alignment-based and causality-based metrics.

Overview of XAI Benchmark for Visual Explanation

Examples of images and human explanation annotations from our published dataset collection for four selected datasets.

Citation

Please consider cite us:

Zhang, Yifei, et al. “XAI Benchmark for Visual Explanation.” arXiv preprint arXiv:2310.08537 (2023).

BibTex:

@misc{zhang2023xai,
      title={XAI Benchmark for Visual Explanation}, 
      author={Yifei Zhang and Siyi Gu and James Song and Bo Pan and Liang Zhao},
      year={2023},
      eprint={2310.08537},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}