Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

XEdgeAI: A Human-centered Industrial Inspection Framework with Data-centric Explainable Edge AI Approach

Truong Thanh Hung Nguyen, Phuc Truong Loc Nguyen, Hung Cao

TL;DR

A novel XAI-integrated Visual Quality Inspection framework that optimizes the deployment of semantic segmentation models on low-resource edge devices and paves the way for the broader adoption of reliable and interpretable AI tools in critical industrial applications.

Abstract

Recent advancements in deep learning have significantly improved visual quality inspection and predictive maintenance within industrial settings. However, deploying these technologies on low-resource edge devices poses substantial challenges due to their high computational demands and the inherent complexity of Explainable AI (XAI) methods. This paper addresses these challenges by introducing a novel XAI-integrated Visual Quality Inspection framework that optimizes the deployment of semantic segmentation models on low-resource edge devices. Our framework incorporates XAI and the Large Vision Language Model to deliver human-centered interpretability through visual and textual explanations to end-users. This is crucial for end-user trust and model interpretability. We outline a comprehensive methodology consisting of six fundamental modules: base model fine-tuning, XAI-based explanation generation, evaluation of XAI approaches, XAI-guided data augmentation, development of an edge-compatible model, and the generation of understandable visual and textual explanations. Through XAI-guided data augmentation, the enhanced model incorporating domain expert knowledge with visual and textual explanations is successfully deployed on mobile devices to support end-users in real-world scenarios. Experimental results showcase the effectiveness of the proposed framework, with the mobile model achieving competitive accuracy while significantly reducing model size. This approach paves the way for the broader adoption of reliable and interpretable AI tools in critical industrial applications, where decisions must be both rapid and justifiable. Our code for this work can be found at https://github.com/Analytics-Everywhere-Lab/vqixai.

XEdgeAI: A Human-centered Industrial Inspection Framework with Data-centric Explainable Edge AI Approach

TL;DR

A novel XAI-integrated Visual Quality Inspection framework that optimizes the deployment of semantic segmentation models on low-resource edge devices and paves the way for the broader adoption of reliable and interpretable AI tools in critical industrial applications.

Abstract

Recent advancements in deep learning have significantly improved visual quality inspection and predictive maintenance within industrial settings. However, deploying these technologies on low-resource edge devices poses substantial challenges due to their high computational demands and the inherent complexity of Explainable AI (XAI) methods. This paper addresses these challenges by introducing a novel XAI-integrated Visual Quality Inspection framework that optimizes the deployment of semantic segmentation models on low-resource edge devices. Our framework incorporates XAI and the Large Vision Language Model to deliver human-centered interpretability through visual and textual explanations to end-users. This is crucial for end-user trust and model interpretability. We outline a comprehensive methodology consisting of six fundamental modules: base model fine-tuning, XAI-based explanation generation, evaluation of XAI approaches, XAI-guided data augmentation, development of an edge-compatible model, and the generation of understandable visual and textual explanations. Through XAI-guided data augmentation, the enhanced model incorporating domain expert knowledge with visual and textual explanations is successfully deployed on mobile devices to support end-users in real-world scenarios. Experimental results showcase the effectiveness of the proposed framework, with the mobile model achieving competitive accuracy while significantly reducing model size. This approach paves the way for the broader adoption of reliable and interpretable AI tools in critical industrial applications, where decisions must be both rapid and justifiable. Our code for this work can be found at https://github.com/Analytics-Everywhere-Lab/vqixai.
Paper Structure (58 sections, 3 equations, 20 figures, 6 tables, 1 algorithm)

This paper contains 58 sections, 3 equations, 20 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. XAI Landscape
  4. Modeling Phase
  5. Scope of Interpretability
  6. Explanation Stage
  7. Explanation Capability
  8. Architecture Approachability
  9. Data-centric in XAI
  10. Application Domains
  11. XAI-based Model Improvement Methods
  12. Data augmentation
  13. Feature augmentation
  14. Loss augmentation
  15. Gradient augmentation
  16. ...and 43 more sections

Figures (20)

  • Figure 1: The XAI landscape, categorizing XAI techniques based on explanation scope, data type, modeling phase, architecture, and application domains.
  • Figure 2: Local Post-hoc XAI Methods for the Semantic Segmentation.
  • Figure 3: The methodology of the XAI-integrated Visual Quality Inspection framework integrated with XAI methods with 6 building modules: (1) Base Model Finetuning with a provided visual quality dataset, (2) Base Model Explanation with XAI, (3) XAI Evaluation, (4) XAI-guided Data Augmentation to improve the base model performance, (5) Edge Model Development on mobile devices and (6) Saliency and textual Explanation for the Edge. The end-users interact with the framework via a mobile application, while the domain experts can interact via a web application.
  • Figure 4: The implementation details of our proposed XAI-integrated Visual Quality Inspection framework.
  • Figure 5: The architecture of the DeepLabv3Plus model, featuring an encoder with Atrous Spatial Pyramid Pooling and a decoder for upsampling and refining segmentation outputs.
  • ...and 15 more figures