Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/sadegh15khedry/handwritten-digit-image-classification
This repository contains code related to the MINST dataset, which is a large database of handwritten digits commonly used for training various image processing systems.
https://github.com/sadegh15khedry/handwritten-digit-image-classification
convolutional-neural-networks deep-learning keras minst neural-networks python
Last synced: about 2 months ago
JSON representation
This repository contains code related to the MINST dataset, which is a large database of handwritten digits commonly used for training various image processing systems.
- Host: GitHub
- URL: https://github.com/sadegh15khedry/handwritten-digit-image-classification
- Owner: sadegh15khedry
- License: mit
- Created: 2021-09-13T12:44:46.000Z (over 3 years ago)
- Default Branch: master
- Last Pushed: 2024-08-08T07:48:16.000Z (5 months ago)
- Last Synced: 2024-08-09T08:35:20.985Z (5 months ago)
- Topics: convolutional-neural-networks, deep-learning, keras, minst, neural-networks, python
- Language: Jupyter Notebook
- Homepage:
- Size: 6.82 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# MNIST Digit Classification
This project implements a neural network model to classify handwritten digits from the MNIST dataset using TensorFlow and Keras. The model architecture includes several dense layers, batch normalization, and dropout for regularization.
- [Directory Structure](#directory-structure)
- [Files and Functions](#files-and-functions)
- [Dataset](#dataset)
- [Model Performance](#model-performance)
- [Installation Guide](#installation-guide)
- [Acknowledgments](#acknowledgments)
- [Further Improvements](#further-improvements)
- [License](#license)
## Directory Structure
```
├── src
│ ├── utils.py
│ ├── model_training.py
│ ├── model_evaluation.py
│ ├── data_preprocessing.py
│ └── data_exploration.py
├── notebooks
│ ├── data_exploration.ipynb
│ ├── data_preprocessing.ipynb
│ ├── model_training.ipynb
│ └── model_evaluation.ipynb
├── models
│ ├── model_custom.joblib
│ └── model_sklearn.joblib
├── environment.yml
└── README.md
```
## Files and Functions
- `utils.py` : Utility functions for various tasks.
- `model_training.py` : Functions for training the model.
- `model_evaluation.py` : Functions for evaluating the model.
- `data_preprocessing.py` : Functions for data preprocessing.
- `data_exploration.py` : Functions for data exploration.
- `data_exploration.ipynb`: Notebook for data exploration.
- `data_preprocessing.ipynb`: Notebook for data preprocessing.
- `model_training.ipynb`: Notebook for model training.
- `model_evaluation.ipynb`: Notebook for model evaluation.
## Dataset
The MNIST dataset is used in this project. It consists of 60,000 training images and 10,000 testing images of handwritten digits (0-9).
## Model Performance
### Train and Validation data Results
- train loss: 0.0616
- train accuracy: 0.9844
- val_loss: 0.0666
- val_accuracy: 0.9859
*Train accuracy and loss*
# Test Data Results
- Test accuracy: 0.9872307777404785
- Test loss: 0.06959925591945648
- precision: 0.9872685356098663
- recall: 0.9872307985515533
- class based reesults
```
precision recall f1-score support
0 1.00 0.98 0.99 1035
1 0.99 1.00 0.99 1181
2 0.99 0.98 0.99 1048
3 0.99 0.99 0.99 1071
4 0.98 0.99 0.98 1023
5 0.98 0.99 0.99 946
6 0.98 1.00 0.99 1031
7 0.99 0.99 0.99 1093
8 0.99 0.98 0.98 1023
9 0.99 0.98 0.98 1043
accuracy 0.99 10494
macro avg 0.99 0.99 0.99 10494
weighted avg 0.99 0.99 0.99 10494
```
*Confiusion matrix*
## Installation Guide
To set up the project environment, use the `environment.yml` file to create a conda environment.
1. **Clone the repository:**
```bash
git clone https://github.com/sadegh15khedry/Handwritten-Digit-Image-Classification.git
cd Handwritten-Digit-Image-Classification
```
2. **Create the conda environment:**
```bash
conda env create -f environment.yml
```
3. **Activate the conda environment:**
```bash
conda activate minst
```
4. **Verify the installation:**
```bash
python --version
```
## Acknowledgments
- Special thanks to the developers and contributors the libraries used in this project, including NumPy, pandas, scikit-learn, Seaborn, Matplotlib and tensorFlow.
- Huge thaks to contributors of the MINST Dataset.
## Further Improvements
- Experimenting models like VGG, ResNet, Inception and etc
# License
This project is licensed under the MIT License - see the LICENSE file for details.