Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/amiriiw/gesture_recognition
Welcome to the Gesture Recognition Project! This project is designed to train a model to recognize hand gestures from images, and then use that model to detect gestures in real-time.
https://github.com/amiriiw/gesture_recognition
gesture-recognition json keras matplotlib numpy pathlib tensorflow
Last synced: 5 days ago
JSON representation
Welcome to the Gesture Recognition Project! This project is designed to train a model to recognize hand gestures from images, and then use that model to detect gestures in real-time.
- Host: GitHub
- URL: https://github.com/amiriiw/gesture_recognition
- Owner: amiriiw
- License: mit
- Created: 2024-08-19T08:51:18.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2024-08-19T08:59:44.000Z (3 months ago)
- Last Synced: 2024-10-11T08:22:40.131Z (28 days ago)
- Topics: gesture-recognition, json, keras, matplotlib, numpy, pathlib, tensorflow
- Language: Python
- Homepage:
- Size: 11.7 KB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Gesture Recognition Project
Welcome to the **Gesture Recognition Project**! This project is designed to train a model to recognize hand gestures from images, and then use that model to detect gestures in real-time.
## Overview
This project consists of two main components:
1. **gesture_recognition_model_trainer.py**: This script is responsible for training a model to recognize hand gestures using an image dataset.
2. **gesture_recognition_detector.py**: This script uses the trained model to detect gestures from images.
## Libraries Used
The following libraries are used in this project:
- **[json](https://pypi-json.readthedocs.io/en/latest/)**: Used for saving and loading class names in JSON format.
- **[pathlib](https://docs.python.org/3/library/pathlib.html)**: Used for working with file paths.
- **[numpy](https://numpy.org/devdocs/user/absolute_beginners.html)**: Used for numerical operations and data manipulation.
- **[tensorflow](https://www.tensorflow.org/)**: Used for building, training, and using the model.
- **[matplotlib](https://matplotlib.org/stable/index.html)**: Used for plotting training history.
## Detailed Explanation
### `gesture_recognition_model_trainer.py`
This script is the core of the project, responsible for training the gesture recognition model. The key components of the script are:
- **ImageDataLoader Class**: Handles loading and preprocessing the image data. The main methods include:
- `load_data()`: Loads training and validation datasets from a directory, applies caching, shuffling, and prefetching to optimize performance.
- **ImageClassifierModel Class**: Handles building, training, and saving the model. The main methods include:
- `build_model()`: Constructs a Sequential model using layers like Conv2D, MaxPooling2D, Dense, and optional data augmentation.
- `train_model()`: Trains the model using the provided datasets and returns the training history.
- `save_model()`: Saves the trained model and class names to files.
- `plot_training_history()`: Plots the training and validation accuracy and loss over epochs.
### `gesture_recognition_detector.py`
This script uses the trained model to detect gestures from images. The key components of the script are:
- **ImageClassifier Class**: Handles loading the model, making predictions, and returning results. The main methods include:
- `predict_image()`: Loads an image, processes it, and predicts the gesture class with confidence.
### How It Works
1. **Model Training**:
- The `gesture_recognition_model_trainer.py` script reads images from the dataset directory.
- The images are resized and prepared for training.
- A Convolutional Neural Network (CNN) model is trained on the dataset to recognize gestures, and the model is saved for later use.
2. **Gesture Detection**:
- The `gesture_recognition_detector.py` script loads the trained model and uses it to predict the gesture class of a given image.
### Dataset
The dataset used for training the model can be accessed via this [Dataset](https://drive.google.com/drive/folders/1R08P48cetCRyVQkAMqE-noCEnF6RTF6t?usp=sharing)
### Installation and Setup
To use this project, follow these steps:
1. Clone the repository:
```bash
git clone https://github.com/amiriiw/gesture_recognition
cd gesture_recognition
```
2. Install the required libraries:
```bash
pip install numpy tensorflow matplotlib
```
3. Prepare your dataset directory as described above.
4. Run the model training script:
```bash
python gesture_recognition_model_trainer.py
```
5. Use the trained model for gesture detection:
```bash
python gesture_recognition_detector.py
```
## License
This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.