Usage Step 1. Install libraries
conda env create -f environment.yml
Step 2. Download the original fer2013 csv file from this link
https://www.kaggle.com/competitions/challenges-in-representation-learning-facial-expression-recognition-challenge/data
Step 3. Put the fer2013 csv file in the main folder
Step 4. Download the pre-trained models from this link
https://1drv.ms/u/s!AmeTT2EpSz40hFVFOvh-r8aSXPlL?e=pFEasy
Step 5. Put the pre-trained models in the following folder
pre_trained_models
To train the model from scratch, run the following
For baseline model: python capsulenet_baseline.py
For batch_norm model: python capsulenet_baseline+batch_norm.py
For 2_conv_layers model: python capsulenet_2_conv_layers.py
For 3_conv_layers model: python capsulenet_3_conv_layers.py
After training the new model will be saved at the result folder
For evaluation on the pre-trained models, run the following
For baseline model: python capsulenet_baseline.py -t -w pre_trained_models/trained_baseline_model.h5
For batch_norm model: python capsulenet_baseline+batch_norm.py -t -w pre_trained_models/trained_baseline+batch_norm_model.h5
For 2_conv_layers model: python capsulenet_2_conv_layers.py -t -w pre_trained_models/trained_2conv_model.h
For 3_conv_layers model: python capsulenet_3_conv_layers.py -t -w pre_trained_models/trained_3conv_model.h5
For evaluation on your model, run the following
python "your_model".py -t -w result/"your_model h5 file"