Facial Beauty Rating and Facial Landmark Detection Implementation on Goblin Real-Time Filtering

Built an end-to-end Goblin filtering system. Facial landmark include points around the eyes, nose, and mouth on any face and are used in many applications, from facial tracking to emotion recognition. The completed project takes in any image containing faces and identifies the location of each face and their facial keypoints.

Quick Start for Windows

1. Download the repository, and navigate to the downloaded folder.

   • Download ZIP or git clone this repository
   • unzip Goblin-Filter-Implementation-main.zip
   • cd Goblin-Filter-Implementation-main

2. Download the datasets.

   • go to Facial Beauty Rating Dataset
   • Download and unzip archive.zip under Goblin-Filter-Implementation-main/data/
   • go to 300 Faces In-the-Wild ibug Dataset
   • Download and unzip ibug_300W_large_face_landmark_dataset.tar.gz under Goblin-Filter-Implementation-main/data/

3. Install a few required pip packages (including OpenCV).

pip install -r requirements.txt

4. Train the model by yourself or skip.

cd src
python facial_beauty_regressor.py
python face_landmark_detection.py

5. Demonstrate the real time filtering with the well-trained model.

python demo.py

• press q to quit the program
• press w to previous filter
• press e to next filter