CarSight 🚗💡

Used Cars Price and Model Assessment in the Pakistani Market

CarSight is an intelligent mobile application designed to transform the used car market in Pakistan by utilizing cutting-edge machine learning and image recognition technologies. The project provides accurate price assessments by analyzing vehicle characteristics and market trends, addressing inconsistencies and subjectivity in traditional pricing methods.

---

Table of Contents

• Features
• Technology Stack
• Dataset
• Model Performance
• Installation and Setup
• Demo Video
• Future Scope
• Contact

---

Features 🌟

1. Image-Based Vehicle Identification
   • Uses YOLOv8 to identify a vehicle’s make, model, variant, year, and condition from uploaded images.
2. Odometer Reading Detection
   • Utilizes EasyOCR integrated with YOLOv8 for accurate mileage extraction.
3. Localized Price Estimation
   • Employs XGBoost and LightGBM models tailored for the Pakistani market, achieving high accuracy in price prediction.
4. Damage Detection
   • Identifies scratches, dents, and other damages on vehicles for a comprehensive assessment.
5. User-Friendly Interface
   • Provides a seamless experience for uploading images, viewing vehicle details, and obtaining instant price estimates.

---

Technology Stack 🛠️

Frontend

• Flutter: Cross-platform development for Android and iOS.

Backend

• Flask: API for seamless integration with machine learning models.
• AWS: Hosting backend services for scalability and reliability.

Machine Learning

• YOLOv8: Vehicle identification, odometer detection, and damage assessment.
• EasyOCR: Text extraction for odometer readings.
• XGBoost & LightGBM: Price prediction based on car details and market trends.

---

Dataset 📊

• Vehicle Identification: 12,000 images for make, model, variant, and year detection.
• Odometer Region Detection: 1,250 images for mileage extraction.
• Damage Detection: 9,500 images for identifying vehicle damage.
• Price Prediction: 21,000 records, including car specifications, historical prices, and market data.

---

Model Performance 📈

Vehicle Identification

• Precision: 90%
• mAP@0.5: 94.9%

Odometer Detection

• Precision: 94%
• mAP@0.5: 94.9%

Damage Detection

• Precision: 93.4%
• mAP@0.5: 90.5%

Price Prediction

• XGBoost R-Squared: 0.98
• Mean Absolute Error (MAE): Lowest in all tests.

---

Installation and Setup 🚀

Prerequisites

• Python 3.8+
• Node.js (for Flutter)
• Firebase CLI

Step 1: Clone the Repository

git clone https://github.com/ahmedshahwar/carsight.git  
cd carsight

Step 2: Install dependencies

pip install -r requirements.txt

Step 3: Run the app

python app.py

---

Demo Video 🎥

Watch the demonstration of CarSight in action:
Click here to view the demo video

---

Future Scope 🔮

• Market Demand Analysis: Incorporate real-time demand and supply trends.
• Integration: Collaborate with financing, insurance, and maintenance services.
• Expanded Features: Include real-time pricing updates and market statistics.

---

Contact 📧

For questions or contributions, reach out:

• Ahmed Shahwar: ahmedshahwarr@gmail.com
• Aymen Zahid: aymenzahid12@gmail.com

---

🚀 Redefining the Used Car Market in Pakistan with AI!