FaceLandmark

Live Video Facial Landmarks Detection Project using MediaPipe - Precursor for Future Projects

This Python project uses MediaPipe to perform real-time face landmark detection and visualize facial blendshapes from webcam video input. It leverages the face_landmarker_v2_with_blendshapes.task model to track facial landmarks and compute blendshape scores for animation or expression analysis.

---

Features

• Real-time face landmark detection with MediaPipe's FaceLandmarker
• Visualizes face mesh, contours, and irises
• Plots facial blendshape scores as a horizontal bar chart
• Easy to run on webcam video input (can be modified for video files)

---

Requirements

• Python 3.8+
• OpenCV
• NumPy
• Matplotlib
• MediaPipe with Tasks API

Install the required Python packages:

pip install opencv-python mediapipe matplotlib numpy

---

Setup

1. Download the Face Landmarker Model

This project requires the face_landmarker_v2_with_blendshapes.task model file, which is not included in this repository.

You can download it from the official MediaPipe documentation:

👉 Download face_landmarker_v2_with_blendshapes.task

2. Place the Model in Your Project

Make sure the model file is saved in your project directory and named:

face_landmarker_v2_with_blendshapes.task

---

How to Run

Once dependencies are installed and the model file is in place, run the program with:

python face_landmark_live.py

Note: Replace face_landmark_live.py with your actual script filename if different.

Press q in the OpenCV window to quit the application.

---

What It Does

• Captures frames from your webcam
• Converts frames to RGB and runs face landmark detection
• Annotates the frame with:
  • Mesh tessellation
  • Facial contours
  • Iris landmarks
• Optionally, plots a blendshape bar chart (currently disabled in live mode for performance)

---

What It Should Look Like

Screenshot 1	Screenshot 2	Screenshot 3

---

Customization Tips

• To use a video file instead of webcam input, change this line:

cap = cv2.VideoCapture(0)

to:

cap = cv2.VideoCapture('path_to_your_video.mp4')

• To enable blendshape bar plots per frame (may slow down performance), you can insert this inside the video loop:

plot_face_blendshapes_bar_graph(detection_result.face_blendshapes[0])

---

License

This project uses MediaPipe, which is licensed under the Apache 2.0 License.