AutoDamageNet – Vehicle Damage Detection

🎥 Live Demo

Check out the live demo of AutoDamageNet deployed on Streamlit:

👉 AutoDamageNet Live App

🚗 Overview

AutoDamageNet is a deep learning project designed to automatically detect vehicle damage. The model classifies images of vehicles into 6 categories:

• Front Breakage
• Front Crush
• Front Normal
• Rear Breakage
• Rear Crush
• Rear Normal

This system can help in insurance claims, accident assessment, and vehicle inspection automation.

---

⚙️ Features

• End-to-end vehicle damage detection.
• Multiple deep learning approaches explored:
  • Baseline CNN
  • CNN with Regularization
  • Transfer Learning with EfficientNet and ResNet
• Finalized with ResNet, achieving ~80% accuracy.
• Built with PyTorch, scikit-learn, and Streamlit (for deployment).

---

🛠️ Tech Stack

• Deep Learning Framework: PyTorch
• Evaluation & Preprocessing: scikit-learn, NumPy, Pandas
• Visualization: Matplotlib, Seaborn
• Deployment: Streamlit

---

Install Dependencies

• pip install -r requirements.txt

Run Streamlit App

• streamlit run app.py