Car-logo-classification

An image classification model from data collection, cleaning, model training, deployment and API integration.
The model can classify 20 different types of logos of cars
The types are following:

• Audi car logo
• BMW car logo
• Chevrolet car logo
• Dodge car logo
• Ford car logo
• Honda car logo
• Hyundai car logo
• Jaguar car logo
• Jeep car logo
• Kia car logo
• Land Rover car logo
• Lexus car logo
• Mazda car logo
• Mercedes-Benz car logo
• Nissan car logo
• Porsche car logo
• Subaru car logo
• Tesla car logo
• Toyota car logo
• Volkswagen car logo

📌 Project Workflow

1️⃣ Problem Statement

Develop an AI-powered model to classify various car logos from images.

2️⃣ Data Collection

• Used DuckDuckGo image search to collect images of 20 car logo categories.
• Automated image downloading with FastAI’s search_images_ddg function.
• Maintained a balanced dataset across all categories.

3️⃣ Data Preprocessing

• Removed duplicate and irrelevant images.
• Standardized image size to 224x224 pixels.
• Applied data augmentation (rotation, flipping, zoom, lighting adjustments) for robustness.

4️⃣ Model Training

• Used ResNet34 with transfer learning.
• Implemented FastAI’s cnn_learner for efficient training.
• Employed cross-entropy loss and Adam optimizer.
• Split dataset into training and validation sets.

5️⃣ Model Evaluation

• Achieved 93% accuracy on the validation set.
• Analyzed misclassifications with a confusion matrix.
• Identified areas for improvement.

6️⃣ Model Deployment

• Exported model using learner.export().
• Built an interactive Gradio-based web interface.
• Deployed the model on Hugging Face Spaces.

7️⃣ Future Improvements

• Expand dataset for better generalization.
• Experiment with ResNet50 or EfficientNet for higher accuracy.
• Implement active learning for dataset refinement.

Data Collection and Preparation

Data Collection

The dataset was collected from DuckDuckGo using car brand names as search terms, ensuring diverse images for training.

DataLoader

The fastai DataBlock API was used to set up the DataLoader, organizing the data into training and validation sets.

Data Augmentation

fastai provides default GPU-accelerated data augmentations, improving the model’s generalization.

For more details, check the Data_prep_logo_classi.ipynb notebook.

Training, Data Cleaning, and Inference

Training

The model was fine-tuned using ResNet101 for 15 epochs, repeated 3 times, achieving an accuracy of approximately 94%. Details can be found in training.ipynb

Data Cleaning

Data cleaning was the most time-consuming part of the process. Since the data was collected from a browser, there was a significant amount of noise, and some images contained irrelevant content. To clean and update the dataset, I used the fastai ImageClassifierCleaner. Data cleaning was performed after each training or fine-tuning session, except for the final iteration, which represented the final model. Details can be found in data_cleaning.ipynb

Inference

For inference, the trained model is used to classify new car logo images. The fastai Learner was utilized to load the fine-tuned model and make predictions. This allows the model to classify unseen car logos accurately, based on the learned patterns during training. Details can be found in inference.ipynb

Model Deployment

I deployed to model to HuggingFace Spaces Gradio App. The implementation can be found in deployment folder or here.

API integration with GitHub Pages

The deployed model API is integrated here in GitHub Pages Website. Implementation and other details can be found in docs folder.

🚀 Project Usage Guide

Follow these steps to set up and run the project on your local machine.

1️⃣ Clone the Repository

git clone https://github.com/fayez94/Car-logo-classification.git
cd Car_Logo_Classification

2️⃣ Create and Activate a Virtual Environment

🔹 For Windows:

python -m venv myvenv
myvenv\Scripts\activate

🔹 For macOS/Linux:

python3 -m venv myvenv
source myvenv/bin/activate

3️⃣ Install Required Dependencies

pip install -r requirements.txt

4️⃣ Run the app file

python app.py

5️⃣ Deactivating the Virtual Environment

deactivate

Contributions

Contributions are always welcome! If you’d like to contribute to the Car Logo Classifier project, here’s how you can help:

How to Contribute:

1. Fork the repository
2. Create a new branch (git checkout -b feature-name)
3. Make your changes
4. Commit your changes (git commit -m 'Add new feature')
5. Push to the branch (git push origin feature-name)
6. Open a Pull Request

Ways You Can Contribute:

• Add support for new car logo categories or labels
• Improve the accuracy of the model by fine-tuning or experimenting with new architectures
• Enhance data collection or augmentation techniques
• Update documentation or add more examples
• Report bugs or suggest new features

Thank you for considering contributing! 🙏

📬 Contact

For any questions or suggestions, feel free to reach out!

📧 Email: mdfayezullah2624@gmail.com
🐙 GitHub: fayez94

This site is open source. Improve this page.