🚗 Car Damage Classification

Classifying the severity of car damage is crucial for insurance claims, safety evaluations, and automated vehicle assessments.
In this project, we compare four different deep learning models to classify car damage into three categories:

• 🟢 Minor
• 🟡 Moderate
• 🔴 Severe

---

📊 Dataset

Source: Kaggle - Car Damage Severity Dataset

The dataset is well-balanced with samples across three classes and is used to evaluate performance across all four models.

📈 Dataset Distribution

---

🧪 Model Comparisons

We applied and evaluated the following models:

---

🔧 1. CNN Model (Built from Scratch)

📉 Training vs Validation Loss

📈 Training vs Validation Accuracy

🧩 Confusion Matrix

🧾 Classification Report

---

📱 2. MobileNetV2 (Transfer Learning)

🏗️ Model Architecture

📉 Loss Curve

📈 Accuracy Curve

🧩 Confusion Matrix

🧾 Classification Report

---

🏛️ 3. VGG16 (Transfer Learning)

🏗️ Model Architecture

📉 Loss Curve

📈 Accuracy Curve

🧩 Confusion Matrix

🧾 Classification Report

---

🧠 Conclusion

This project provides a comprehensive comparison between custom CNN and popular pretrained architectures (MobileNetV2, VGG16).
It reveals:

• 📉 Training loss trends
• 📈 Accuracy performance
• 🧩 Confusion matrices
• 📝 Detailed classification reports

Each model has strengths, and the choice depends on the deployment constraints and accuracy requirements.

---

🚀 Future Work

• ✅ Add more advanced architectures (e.g., EfficientNet, ResNet)
• 🧪 Integrate cross-validation
• 💡 Deploy as a web app for user upload and real-time predictions

---

🛠️ Tech Stack

• Python 🐍
• TensorFlow / Keras
• OpenCV
• Scikit-learn
• Matplotlib / Seaborn
• NumPy / Pandas

🛠️ Getting Started

📦 Prerequisites

Make sure you have the following installed:

• Python
• Git
• Jupyter Notebook or JupyterLab
• A modern GPU (recommended for training)
• pip or conda for package management

🧰 Installation

Clone the repository:

git clone https://github.com/your-username/car-damage-classification.git
cd car-damage-classification