SegClarity:一个基于归属的XAI工作流程,用于评估历史文档布局模型.
Iheb Brini1,2, Najoua Rahal2, Maroua Mehri1
1Ecole Nationale d'Ingénieurs de Sousse, Laboratory of Advanced Technology and Intelligent Systems (LATIS), Université de Sousse, Sousse 4054, Tunisia.
Journal of imaging
|December 24, 2025
概括
这项研究引入了SegClarity,一个新的工作流程和指标 (归因一致性得分),以提高用于历史文档分割的深度学习模型的可解释性,提高对AI输出的信任.
科学领域:
- 计算机视觉 计算机视觉
- 人工智能的人工智能
- 数字人文学科 数字人文学科
背景情况:
- 深度学习在历史文档细分方面表现出色,但缺乏可解释性,阻碍了信任.
- 可解释的人工智能 (XAI) 技术产生热图,但缺乏对评估指标的共识.
- 现有的XAI方法经常产生类似于细分输出的地图,限制了可解释性.
研究的目的:
- 介绍 SegClarity,一个新的工作流程,用于将可解释性整合到历史文档语义细分中.
- 引入归因一致性得分 (ACS) 用于对XAI归因图进行定量评估.
- 在历史文档分析中提高深度学习模型的透明度和可靠性.
主要方法:
- 开发了SegClarity工作流程,将细分任务的视觉和定量评估结合起来.
- 引入了归因一致性得分 (ACS) 作为一种新的可解释性指标.
- 使用U-Net模型和四种XAI方法对历史文档数据集进行了广泛的实验,包括与RISE和MiSuRe进行比较.
主要成果:
- SegClarity 工作流显著提高了用于历史文档分割的深度学习模型的可解释性和可靠性.
- 广泛的定性和定量评估证明了拟议方法的有效性.
- 工作流显示了Cityscapes数据集上的概括能力,用于城市场景细分.
结论:
- SegClarity提供了一个强大的框架,用于评估XAI在文档图像分析.
- 该ACS指标提供了对归因地图可靠性的定量见解.
- 该研究增强了人工智能驱动的历史文档分析的信任和可重现性.
相关概念视频
Methods of Documentation VI: Case Management Model
833
The case management model is a multidisciplinary approach that involves healthcare professionals from diverse disciplines, such as physicians, nurses, therapists, social workers, and pharmacists, working collaboratively to address the various needs of patients. Each healthcare professional brings unique expertise and perspectives, contributing to a more comprehensive understanding of the patient's condition and tailoring treatment plans accordingly.
For example, a patient with a chronic...
For example, a patient with a chronic...
833
Methods of Documentation IV: Focus Charting
1.6K
Focus Charting, also known as the focus charting system or "focus documentation," is a systematic documentation approach used in healthcare to organize patient information in medical records.
It typically involves three columns for recording information:
It typically involves three columns for recording information:
1.6K
Methods of Documentation III: PIE
2.0K
Problem-intervention-evaluation (PIE) is a systematic approach to documentation used in healthcare settings for clinical decision-making and patient care planning. It is a structured approach to organizing patient data based on problems, interventions, and evaluations. Here's a breakdown of its key features and considerations:
2.0K
Methods of Documentation I: Source-Oriented Records
1.7K
Source-oriented records, or SOR, are medical record-keeping organized by the data source. The SOR system was first developed in the mid-1900s to organize the growing patient data in hospitals and other healthcare facilities.
In an SOR, each discipline involved in patient care maintains a separate medical record section. This record-keeping method enables easy tracking of patient progress and ensures healthcare staff have access to up-to-date information.
Key Attributes include the following:
In an SOR, each discipline involved in patient care maintains a separate medical record section. This record-keeping method enables easy tracking of patient progress and ensures healthcare staff have access to up-to-date information.
Key Attributes include the following:
1.7K
Methods of Documentation II: POMR
1.3K
The Problem-Oriented Medical Record (POMR) revolutionized medical record-keeping by introducing a systematic approach focusing on the patient's problems rather than merely listing symptoms. Dr. Lawrence Weed's introduction of this method in the 1960s marked a significant advancement in medical documentation. The POMR framework consists of four key components: the database, problem list, plan of care, and progress notes.
1.3K
Models, Theories, and Laws
8.2K
Scientists frequently use models to help them comprehend a specific collection of phenomena. In physics, a model is a condensed version of a physical system that is too complex to study thoroughly. One such example is the light wave model; unlike water waves, light waves are typically invisible to us. Nonetheless, it is helpful to think of light as being composed of waves, since investigations show that light behaves like water waves. Since it is impossible to visually see what is genuinely...
8.2K

