AI-Driven Component-Level Vehicle Damage Analysis Across Diverse Types and Challenging Image Conditions
Youngsun Cho, Nakyung Lee, Nakyoung Kim, Seongbeom Kwak, Minsoo Jeong, Hyerin Chung, Han-Sol Lee, Jihwan Woo · 2025
This paper presents an AI-driven vehicle damage analysis system that offers precise assessment across various vehicle categories, such as sedans, SUVs, and trucks, under diverse image conditions. The system integrates advanced image quality assessment, filtering, and enhancement techniques, followed by a component-level damage analysis. A core feature is its continuous learning capability, allowing adaptation to new vehicle models and damage patterns. The system's robustness is validated through experiments involving state-of-the-art models, and its practicality is supported by user studies demonstrating the benefits of functionalities like similar quote estimation and mobile deployment. These advancements enhance damage analysis accuracy and user experience, showcasing its potential in the insurance and consumer electronics service sectors.