Hello, My name is Sarvesh Shanmugam current third year college student majoring in CS and minoring in business. Have an idea and would love to connect with a mentor through gsoc-2025!
Car Damage Detector idea(DAMAGE DOCTOR)
Project Overview
The Car Damage Detector is an AI-powered iOS application designed to analyze vehicle damage and estimate repair costs in real time. By leveraging deep learning and computer vision, the app will provide users with accurate assessments and visual explanations of detected damage. This tool will assist insurance companies, car rental services, and individual vehicle owners in making informed decisions about repairs.
Problem Statement
Assessing car damage manually can be subjective, time-consuming, and prone to inconsistencies. Current solutions lack automation, require expert evaluation, or fail to provide real-time cost estimation. This project aims to address these issues by developing an AI-driven mobile application that can analyze damage instantly and provide reliable repair cost estimates.
Solution Approach
- Model Training & Optimization
- Train a deep learning model using TensorFlow and Scikit-learn with a dataset of vehicle damage images.
- Achieve optimal classification accuracy through model fine-tuning.
- Improve feature extraction and model interpretability using OpenCV and Plotly.
- iOS Application Development
- Develop the app using Swift and SwiftUI for an intuitive user interface.
- Integrate the trained model using Core ML for efficient on-device inference.
- Utilize Vision Framework and AVFoundation for image/video analysis.
- Display damage classification results with AI-powered explanations and saliency map visualizations.
- Cost Estimation & Reporting
- Implement a cost prediction system based on detected damage severity.
- Generate detailed reports for users, including suggested repair costs and severity levels.
Technologies Used
- Machine Learning & AI: TensorFlow, Keras, Scikit-learn
- Computer Vision: OpenCV, Plotly
- iOS Development: Swift, SwiftUI, Core ML, Vision Framework, AVFoundation
- Deployment: On-device ML model integration using Core ML
Expected Deliverables
- A trained deep learning model optimized for real-time damage classification.
- A fully functional Swift-based iOS application with:
- Image and video-based damage detection.
- AI-generated damage severity insights and cost estimation.
- Interactive saliency map visualizations.
- A final report summarizing model performance and application usability.
Impact & Benefits
- Efficiency: Instant damage assessment reduces manual inspection time.
- Cost Savings: Helps users avoid overpaying for repairs.
- Scalability: Can be adopted by insurance companies, car rental services, and individuals.
This project will push the boundaries of AI-driven vehicle assessment, bringing automation and accuracy to car damage evaluation.