Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/sharan-vj/end2end_bear_panda_binary_image_classification
End to End Deep learning project, Fine tuned VGG19 model to classify bears and pandas, deployed via a web interface for quick and easy predictions.
https://github.com/sharan-vj/end2end_bear_panda_binary_image_classification
binary-classification cicd deep-learning docker end-to-end-project flask github-actions mlops-project transfer-learning vgg19
Last synced: about 2 months ago
JSON representation
End to End Deep learning project, Fine tuned VGG19 model to classify bears and pandas, deployed via a web interface for quick and easy predictions.
- Host: GitHub
- URL: https://github.com/sharan-vj/end2end_bear_panda_binary_image_classification
- Owner: Sharan-vj
- License: mit
- Created: 2024-10-27T11:08:14.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2024-10-27T11:45:41.000Z (3 months ago)
- Last Synced: 2024-12-17T00:13:37.263Z (about 2 months ago)
- Topics: binary-classification, cicd, deep-learning, docker, end-to-end-project, flask, github-actions, mlops-project, transfer-learning, vgg19
- Language: Jupyter Notebook
- Homepage: https://github.com/Sharan-vj/End2End_Bear_Panda_Binary_Image_Classification
- Size: 12.1 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# End2End Bear-Panda Binary Image Classification Using Transfer Learning
### Project Description
This project presents a complete end-to-end machine learning workflow to classify images of bears and pandas using a fine-tuned **VGG19** model. With the goal of accurately distinguishing between these two animals, the model leverages transfer learning to enhance prediction performance. A web-based application is built to allow users to upload images and instantly receive classification results.
### Screenshots
### Technologies Used in This Project
* Python: For model development and backend integration
* TensorFlow/Keras: For implementing and fine-tuning the VGG19 model
* Pandas & NumPy: For data processing and manipulation
* Flask: For deploying the model as a web application
* Jupyter Notebook: For Exploratory Data Analysis (EDA) and model development
* HTML/CSS: For designing the web-based user interface
* Git/GitHub: For version control and project collaboration
* Docker: For containerizing the application
* GitHub Actions: To automate CI/CD workflows
* AWS: For hosting and deployment of the web app
### Model Performance
### Installation and Setup
1. **Clone the Repository**:
```bash
git clone End2End_Bear_Panda_Binary_Image_Classification.git
cd End2End_Bear_Panda_Binary_Image_Classification
```
2. **Create a Virtual Environment**:
```bash
conda create --name python=3.10 -y
```
3. **Activate the Virtual Enviroment**
```bash
conda activate
```
4. **Install the Dependencies**:
```bash
pip install -r requirements.txt
```
5. **Run the Flask Application**:
```bash
python app.py
```
The application will be available at `http://127.0.0.1:8080/`.
### Usage
* Upload Image: Use the web interface to upload an image of a bear or panda.
* Get Prediction: The fine-tuned `VGG19` model will classify the image as either a bear or a panda, displaying the result instantly.
### Dataset
* The dataset used in this project contains labeled images of bears and pandas. The data is stored in the data/ directory. The model is trained on a balanced dataset to ensure consistent performance across both classes.
### Model
* **Model Architecture**: VGG19
* **Fine-tuning**: The final few layers of the VGG19 model were re-trained on the bear vs. panda dataset to enhance accuracy.
* **Model Saving**: The base model is saved as `base_model.h5` and The model trained weights are saved as `model.weights.h5` inside the models/ directory for easy reuse.
### Contributing
* Contributions are welcome! If you have any suggestions or improvements, please raise an issue or submit a pull request.
### License
* This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for more details.
### Acknowledgments
* Special thanks to the open-source community for the libraries used in this project.