emojify me now
Install requirements
python -r requirements.txt
-
model.py
- To train the model
- outputs the model & weight in model directory
- also outputs the graph having the logs of training
- arguments -e/--epochs
- pass the integer value of epochs
-
get_images_from_webcam.py
- to create dataset using webcam (including all the required subfolders)
- It includes face alignment and face detection
- argparse -l/--label
- pass the argument like smile/wink/laugh according to your need
-
utils.py
- contains useful required functions
-
memojifier.py
- It takes the stream of photos and converts the current emotion of the face into animated emoji
Current Network Architecture
Result: with validation accuracy of 80%
Currently model is unstable and overfits easily (:funny)
Try to train the model with as many as possible images and different environment
this project is currently in experimental stage and will be worked upon
Got some issue? raise an issue and if you have something to contribute then push a PR.