VFourCuts(브이네컷)

Overview

VFourCuts is a photo booth application written in Python that performs real-time gesture recognition and automatically captures photos when a V-sign is detected. This program uses OpenCV and Tensorflow for gesture recognition and captures and composes images after a countdown.

Key Points:

1. Hand Gesture Recognition:

• The script uses the MediaPipe library for hand landmark prediction and recognizes the V-sign(peace) gesture in real-time camera frames.

2. Gesture Prediction Model:

• A pre-trained TensorFlow/Keras model for hand gesture recognition is loaded using load_model. The model is designed to predict gestures based on hand landmarks.

3. Camera Capture and Countdown:

• The script captures video frames from the default camera (cv2.VideoCapture(0)).
• A countdown is initiated when the V-sign(peace) gesture is detected, and images are captured every 5 seconds for a total of 4 repetitions.

4. Image Processing:

• Hand landmarks are extracted, and the V-sign(peace) gesture is predicted for each detected hand.
• Landmarks are visualized on the frames using mpDraw.draw_landmarks.
• Images are combined into a single image after a 5-second countdown.

Installation

Supported Python version: 3.11

1. Install Python: Download and install Python from the official Python website Make sure you are using a supported version of Python.
2. Download the sources: Run git clone https://github.com/heoheosilsil/vFourCuts.git or download the zip archive of this project.
3. Install the dependencies: In the project directory, use the command pip install -r requirements.txt to install the dependencies.

Usage

1. Run the main.py file.
2. Once the program starts, gesture detection will start.
3. When a V-Sign is detected, a 5 second countdown will start and a photo will be taken.
4. A total of 4 photos will be composited and saved in a file named result_vfourcuts.png.

Screenshot

When the program is running, you will see:

• A countdown timer
• The current number of detected V-Signs

Result

Dependencies

The project uses OpenCV for image processing and Tensorflow for gesture recognition. A detailed list of dependencies and their versions can be found here.

References

This program uses pre-trained datasets from the following website for machine learning gesture recognition.

• https://techvidvan.com/tutorials/hand-gesture-recognition-tensorflow-opencv/

Limitations:

• The technique requires nearly perfect top-down views for accurate gesture recognition.

Contributions

If you would like to contribute to this project, please open a pull request or issue. All contributions are welcome!