Face-Keypoints

How to Use

• Clone this repo to your machine git clone https://github.com/Rajrup/Face-Keypoints.git
• Install the python dependencies (see below)
• Download the checkpoints(see below)
• Double check the model paths
• Ready to run pipeline.py

Checkpoint Preparation

• Find all the models: Google Drive
• Extract the Tensorflow model files into models/ folder.
  • Face: Path will look this - ./models/frozen_inference_graph_face.pb
  • PRNet: Path will look this - ./models/PRNet/net-data/256_256_resfcn256_weight.data-00000-of-00001

Requirements

This code has been tested in Python 3.7. See requirements.txt for python packages.

pip install -r requirements.txt

Running Pipeline

• Tensorflow Pipeline:

  python pipeline.py

• Tensorflow Serving Pipeline: TODO

• Output in the media folder. Check the file output.avi for keypoints.

Components

One module's output will go to the next one

• Video Reader
• Face Detection (Face)
• Key Point Extraction (PRNet)