pawelkopec/smile-warrior

This repo contains code used for smile detector for project of Gradient Science Club.

smile-warrior

This repo contains code used for smile detector for project of Gradient Science Club.

To launch web app:

py app.py --cascade  haarcascade_frontalface_default.xml --model  model.hdf5 --weight  weights.03.hdf5

To create dataset:

Run:

   python adjust_dataset.py 

Result will be a csv file with dataset for Smile-Warrior (by default in: "smile_warrior_dataset.csv").

To load dataset:

Use:

   python data_server.py

To split whole dataset for Smile-Warrior into respectively: x_train, y_train, x_test, y_test, x_validate and y_validate datasets.

Training logs:

Data from 3 training sessions, accordingly for 10, 20 and 30 epochs (without augmentation)

Accuracy:

Loss:

Validation accuracy:

Validation loss: