aarsri/emogif

EmoGIF

Aarohi Srivastava

Getting Started

Requirements and Dependencies

• MUST HAVE OPENCV 4.5.0
• Python 3 (for Python 3.9, use opencv=4.5.0=py39_2)
• conda install -c conda-forge dlib
• conda install -c conda-forge scipy
• pip install opencv-contrib-python
• pip install h5py
• pip install trimesh==2.37.1
• pip install mls
• pip install scikit-image
• conda install -c kitware-danesfield-cf rtree

You will also need the file shape_predictor_68_face_landmarks.dat. This can be found at https://github.com/davisking/dlib-models. Please put this in the EmoGIF-master directory.

In case something isn't working right, please see the requirements.txt file exported from my personal environment to see the versions I had while writing this code.

Run Without GUI

$python bringToLife.py <path-to-selfie-image> <emotion> <output-filename>

For example:

$ python bringToLife.py input_images/Frida1.jpg peaceful Frida1-peaceful.mp4

emotion must be one of: excited, happy, whimsical, love, peaceful, nostalgic, gloomy, confused

output filename must be of type mp4

Your output video can be found in the video_generated folder.

Run With GUI

The GUI uses Zenipy. Please follow https://github.com/poulp/zenipy if you would like to use this. Once set up, run:

$ python bringToLife_gui.py

Note that Zenipy does not handle typos or backspace in user input.