The goal of this project is to build a model that can recognize the emotion of a face. The model will be trained on a dataset of images of faces and their emotion. The model will be able to recognize the emotion of a face based on the image.

We have used pytorch, pytorch lightning, and pytorch vision.

The file containig the model used, resnet can be found in the dl/base_torch_modules/resnetmodel.py. The pretrained model is contained in the dl/models/resnet/checkpoint.ckpt.

Make sure you have at least python 3.9 installed. If the dlpm folder is empty, you need to clone the repository.

git clone https://github.com/GiulioZani/dlpm

Then install the dependecies.

pip install -r dlpm/requirements.txt

pip install -r dl/requirements.txt

dlpm is a framework for running and testing deep learning models. The code relevant to this specific task is contained in the dl folder.

The dataset can be downloaded form kaggle. The size is 9GB.

python -m dl.preprocess --dataset_path=<unzipped folder containing data> --destination_folder=<preprocessed location>

First you need to edit the dl/models/resnet/default_parameters.json file and set the data_location parameter to the location of the preprocessed dataset. Then run:

python -m dlpm test resnet

To classify webcam videos, run:

python -m dlpm exec resnet