Emotion Classification

• Used transfer learning to build a convolutional neural network with FER2013 dataset that can classify emotion from different photos (Phoebe from F.R.I.E.N.D.S).

Dataset

For Training (FER 2013)

Angry	Disgust	Fear

Happy	Neutral	Sad	Surprise

For Testing (Phoebe)

Predictions

Angry	Surprise

Get Started

• (Optional) Create a virtual environment and activate it.

virtualenv venv
source venv/bin/activate

Installation

• Make sure you have all the packages installed from requirements.txt.

pip install -m requirements-dev.txt

pip install -m requirements.txt

• Use the Jupyter Notebook (test.ipynb) to build the model and test the model's accuracy.

Train the Model

python3 model/train.py

Fine-tune the Model

python3 model/fine_tune.py

Test the Model

pytest -v --cov --cov-report term-missing