Automobile Inspector

Abstract

The used car market is expanding rapidly, yet fraud, misrepresented evaluations, and biased pricing erode trust between buyers and sellers. Traditional inspections are time-intensive, expensive, and subjective. To address these, the Automobile Inspector an AI-powered system automates vehicle condition assessment, fair pricing, and cost estimation, including a trip cost, fuel efficiency, and city-to-city calculator. It leverages Python, TensorFlow, PyTorch, OpenCV, and YOLO with a teacher-student framework (YOLOv8x as teacher, YOLOv8L as student) for detecting components like engines, interiors, and exteriors. The interior model uses DenseNet121 with transfer learning; sound analysis for early mechanical fault detection employs MFCC+CNN; text-based models incorporate XGBoost. Datasets are sourced from IAAI.com and Copart.com for engine sounds, interior/exterior images, and text features, plus additional engine sounds from YouTube and TikTok. Sophisticated image processing identifies dents, scratches, and structural defects. The frontend is built with Flutter and Dart, connected to Firebase for user authentication, while the backend uses FastAPI and MongoDB to store user-inputted images. The system provides real-time market-based price forecasting, a RAG based chatbot for interactive queries, and a report generation feature that produces full-fledged PDF reports of car analysis. Overall accuracy we got in models is 81% ,it delivers objective, data-driven insights that empower buyers with transparency and confidence while boosting sellers' credibility.