A PyTorch Implementation of iCaRL

This is an unofficial implementation of iCaRL: Incremental Classifier and Representation Learning on Mninst, FashionMnist and Kmnist datasets. This code is modified from https://github.com/donlee90/icarl.

Usage

Train the model

1. Clone the project to directory

git clone https://github.com/bill86416/ICaRL.git

2. Initiate the conda environment

conda env create -f environment.yml -n icarl
conda activate icarl

3. Run Mnist experiment

sh run_mnist.sh

3. Run notMnist experiment

sh run_notmnist.sh

3. Run Fashion Mnist experiment

sh run_fashionmnist.sh

Ackowledgement

Please email to Chih-Hsing Ho (Bill) (bill86416@gmail.com) if further issues are encountered.