Awesome

This repository contains the common lists used in Class-Incremental Learning (CIL) works. It is there to provide a fair comparison between the different methods. The lists are available for the following datasets:

• CIFAR-100: 100 classes
• ImageNetSubset: 100 classes
• TinyImageNet: 200 classes
• ILSVRC: 1000 classes
• GoogleLandmarks: 1000 classes
• iNaturalist: 1000 classes

Each folder contains a README.md file that explains how to download the dataset. The lists are provided in the following format: a text file with the path to the images and their labels. The first line of the file indicates the path to the images. The following lines are of the form: path/to/image label. The labels are integers starting from 0.

datasets_mean_std.txt contains the mean and standard deviation of the datasets. The mean and standard deviation are computed on the training set.

A loader is provided in the repository utilsCIL in MyImageFolder.py to load the lists.