/Facial-Keypoint-Detector

Facial Keypoint detection and style filters overlay using CNN

Primary LanguagePython

Overview

This project detects facial keypoints and let's you overlay snapchat style filters to it.

Facial Keypoint Detection Filter Overlay
Facial Keypoint Detection Snapchat Style Overlay

Setup

For better and faster results, GPU support is required. CPU should work fine as well.

  1. Clone the repository, and navigate to the downloaded folder.
git clone https://github.com/ysharc/Facial-Keypoint-Detector.git
cd Facial-Keypoint-Detector

  1. Create (and activate) a new environment with Python 3.6 and the numpy package.

    • Linux or Mac:
    conda create --name FKD-App python=3.6 numpy
source activate FKD-App
    • Windows:
    conda create --name FKD-App python=3.6 numpy scipy
activate FKD-App

  2. Install/Update TensorFlow (for this project, you may use CPU only).

    • Option 1: To install TensorFlow with GPU support, follow the guide to install the necessary NVIDIA software on your system.
    pip install tensorflow-gpu==1.3.0
    • Option 2: To install TensorFlow with CPU support only:
    pip install tensorflow==1.3.0

  3. Install/Update Keras.

pip install keras -U

  1. Switch Keras backend to TensorFlow.

    • Linux or Mac:
    KERAS_BACKEND=tensorflow python -c "from keras import backend"
    • Windows:
    set KERAS_BACKEND=tensorflow
python -c "from keras import backend"

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

pip install -r requirements.txt

Data

All of the data you'll need to train a neural network is in the AIND-CV-FacialKeypoints repo, in the subdirectory data. In this folder are a zipped training and test set of data.

  1. Navigate to the data directory
cd data
  1. Unzip the training and test data (in that same location). If you are in Windows, you can download this data and unzip it by double-clicking the zipped files. In Linux/Mac, you can use the terminal commands below.
unzip training.zip
unzip test.zip

You should be left with two .csv files of the same name. You may delete the zipped files.

Troubleshooting: If you are having trouble unzipping this data, you can download that same training and test data on Kaggle.

Now, with that data unzipped, you should have everything you need!

Quick Start

  • If you're interested in knowing the whole process, then run the interactive jupyter notebook and follow the instructions.
jupyter notebook CV_project.ipynb
  • Web app [coming soon!]