The Vehicle Classification System is an advanced deep learning solution that accurately classifies vehicles into three categories: Truck, Minitruck, and Car. Built using PyTorch and ResNet50, the system provides high accuracy detection through multiple interfaces including a user-friendly GUI, web server, API, and service integrations.
- Production-Ready: Enterprise-grade vehicle classification with 94.5% validation accuracy
- Pre-trained Model: Ready to use without additional training
- Multi-Platform: Works on Windows, Linux, and macOS (service mode Windows-only)
- Multi-class Vehicle Classification: Accurately identify vehicles as Truck, Minitruck, or Car
- Multiple Interface Options:
- Interactive Desktop GUI application
- Web Server with browser interface
- RESTful API for system integration
- Windows Service for continuous operation
- Discord Integration: Real-time logging and notifications
- Feedback Collection System: Continuous model improvement through user feedback
- Batch Processing: Process multiple images at once
- Model Retraining: Update the model with new data through the retraining system
- Python 3.10 or newer
- Windows OS (for service deployment)
- 2GB+ RAM
- CUDA-compatible GPU (optional, for faster inference)
-
Clone the repository:
git clone https://github.com/Xer0bit/Vehicle-Car-MiniTruck-Truck-v2.git cd Vehicle-Car-MiniTruck-Truck-v2 -
Run the setup script:
# For Windows setup.bat # For PowerShell .\setup.ps1 -
Activate the environment:
# For Windows activate_env.bat # For PowerShell & .\vision\Scripts\Activate.ps1
The system can be used in multiple ways:
Run the GUI application for interactive classification:
python classifier_gui.py
Features:
- Drag and drop image support
- Batch processing
- Visual results with confidence scores
- Feedback submission for continuous learning
Start the web server:
python web_server.py
Then access the web interface at http://localhost:5000
The system provides a RESTful API for integration with other applications. See API_README.md for detailed documentation.
Basic example:
import requests
api_url = "http://localhost:5000/api/classify"
api_key = "your_api_key_here"
headers = {"X-API-Key": api_key}
with open("path/to/image.jpg", "rb") as image_file:
files = {"image": image_file}
response = requests.post(api_url, headers=headers, files=files)
result = response.json()
print(f"Classification: {result['class_name']} (Confidence: {result['confidence']:.2f})")To install as a Windows service:
install_service.bat
To start/stop the service:
run_service.bat
stop_service.bat
The system collects feedback during operation, which can be used to continuously improve the model. To retrain with new data:
python retrain.py --data feedback_data/corrected_images/
- System collects misclassified images in
feedback_data/ - User provides corrected labels through GUI or feedback form
- Retrain model using the corrected dataset
- Evaluate new model performance
- Deploy updated model to production
For detailed information about the training process, see our training notebook.
The model is trained on a comprehensive dataset of vehicle images organized in the following structure:
dataset/
train/
bus/ # Mini-truck category images
... # Other vehicle class folders
val/
... # Validation data with same structure as train
test/
... # Test images for evaluation
The dataset includes various vehicle types that are mapped to three main categories:
- Truck: Heavy trucks and large commercial vehicles
- Minitruck: Buses, fire engines, and medium-sized trucks
- Car: Personal vehicles and smaller automobiles
Our ResNet50-based model achieves excellent performance metrics:
| Metric | Train | Validation |
|---|---|---|
| Accuracy | 100% | 94.5% |
| Loss | 0.000097 | 0.003131 |
These results were achieved after 20 epochs of training with the following configuration:
- Learning Rate: 0.1 with StepLR scheduler
- Batch Size: 64
- Image Size: 224x224
- Data Augmentation: Normalization with ImageNet stats
- Core Classification Engine: PyTorch with ResNet50 architecture
- UI Layer: PyQt5 for desktop GUI
- Web Server: Flask for web interface and API
- Service Layer: Windows service integration using NSSM
- Logging System: Multi-level logging with Discord integration
- Feedback System: Data collection for continuous improvement
The pretrained model files are stored in the pt directory:
pt/
model_weights.pth
label_encoder.pth
The system loads these files by default, providing immediate classification capabilities without requiring retraining.
⬇️ Download Pre-trained Model (108MB)
- Base: ResNet50 (pretrained on ImageNet)
- Fine-tuned with transfer learning
- Input size: 224x224 RGB images
- Output: 3 classes (Truck, Minitruck, Car)
For detailed information about specific components:
Contributions are welcome! Please feel free to submit a Pull Request.
This project is licensed under the MIT License - see the LICENSE file for details.
Performance benchmarks on various hardware configurations:
| Hardware | Processing Time (ms/image) | FPS |
|---|---|---|
| NVIDIA RTX 3080 | 8.2 ms | 122 |
| NVIDIA GTX 1660Ti | 12.7 ms | 79 |
| Intel i9-12900K (CPU) | 34.5 ms | 29 |
| Raspberry Pi 4 | 312 ms | 3.2 |
If you use this model in your research or project, please cite:
@software{xer0bit2025vehicle,
author = {Muhammad Sameer Akram},
title = {Vehicle-Car-MiniTruck-Truck Classification System},
year = {2025},
url = {https://github.com/Xer0bit/Vehicle-Car-MiniTruck-Truck-v2},
version = {2.0}
}- ResNet architecture by Microsoft Research
- PyTorch team for the deep learning framework
- Dataset contributors for training data
© 2025 BitLogicX | Vehicle Classification System v2.0 | xer0bit.com