TouchDesigner Multi-face Tracking with OpenCV and YuNet

Overview

This repository provides implementations for multi-face tracking in TouchDesigner using two methods:

• YuNet: A deep learning-based face detection model.
• OpenCV Cascade Classifier: A traditional method using Haar cascades.

---

YuNet

In the folder face_detection_yunet, you will find the TouchDesigner implementation of YuNet.

Video Source

• This example uses op('VideoSource') as the source for face detection.
• To test with your own video, replace the file path in the operator.

Model Setup

• In DAT script2_callbacks, update the system path to the face_detection_yunet folder on line 6.
• Path format is r"C:\.... for Windows.

Hardware Acceleration

• This implementation attempts to use CUDA if OpenCV DNN CUDA support is available.
• If CUDA is not found, it will automatically switch to CPU processing.

---

OpenCV

The folder OpenCV contains an implementation using the Cascade Classifier from OpenCV in TouchDesigner.

Features

• This example detects faces and applies a blur effect on them.

---

Requirements

• TouchDesigner
• OpenCV with DNN module support (for YuNet)
• CUDA (optional, for hardware acceleration)

---

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/TD-Multi-face-Tracking.git

2. Install required Python packages:

   pip install opencv-python opencv-python-headless

3. Update system paths in the scripts as mentioned above.

---

Usage

1. Open the corresponding TouchDesigner project file.
2. Adjust video source paths as needed.
3. Run the project and observe real-time face tracking.

---

License

This project is licensed under the MIT License.

---

Acknowledgments

• OpenCV
• YuNet Face Detection Model