esssyjr/6-places-classification

Our project focuses on leveraging state-of-the-art machine learning techniques, specifically MobileNet transfer learning, to develop a robust image classification model capable of distinguishing between six distinct environmental categories. The targeted classes include 'buildings,' 'forest,' 'glacier,' 'mountain,' 'sea,' and 'street.'

Image Classification Model

This repository contains a deep learning model for classifying images into categories such as 'buildings', 'forest', 'glacier', 'mountain', 'sea', and 'street'. The model, built using MobileNet transfer learning, achieved an 81% training accuracy and 78% validation accuracy.

Computer Vision Work

Table of Contents

1. Dataset
2. Model Architecture
3. Results
4. Future Work
5. Contributing
6. Citation

Dataset

The training dataset consists of a balanced collection of 11,230 images for each class: 'buildings', 'forest', 'glacier', 'mountain', 'sea', and 'street'. A validation set of 598 images and a test set of 7,301 images are also part of the dataset. The data is split into training and validation sets to evaluate the model's performance.

Data Augmentation

I used an image data generator to augment the data, providing more photos to improve the model's robustness.

Model Architecture

The classification model is built on top of the MobileNet architecture, a lightweight and efficient convolutional neural network (CNN) suitable for mobile and embedded vision applications. I excluded the top layer and added my own. I used fine-tuning techniques to achieve better accuracy. The use of different transfer learning models was explored.

Results

The model achieved an impressive 81% accuracy on the training set and 78% accuracy on the validation set.

Future Work

While the current model performs well, there is room for improvement. Future work could focus on:

• Fine-tuning: Further refine the model by adjusting hyperparameters or exploring different architectures.

• Data Expansion: Increase the dataset size with additional diverse images to enhance the model's generalization capability.

• Regularization Techniques: Implement regularization methods to prevent overfitting and improve model performance on unseen data.

• Exploration of Other Techniques: Investigate and implement other state-of-the-art techniques in computer vision for potential performance gains.

Contributors are encouraged to explore these avenues and contribute to the ongoing development of the model.

Contribution

Contributions are welcome! If you find any issues or have suggestions, feel free to open an issue or submit a pull request.

Citation

Dataset Source - Kaggle Intel Image Classification Data