Download data from Kaggle and placed the unzipped data under the data/fer2013 directory of this project.
run the following script under project root
python -m cnn.model_face --model=4 --name=CustomModelName --epoch=65 --task=train,testShould be able to achieve 67% accuracy on test set.
--model=1: A test Network in Network model
--model=2: A NIN model
--model=3: A residual network without bottleneck
--model=4: A residual network with bottleneck
The epoch number to run
A unique name as an identifier for the model. Duplicate name of existing model would override the saved model on disk
Specify tasks to run, default to train,test. If you only want to test an existing model, run with --task=test.