21 datasets found

A novel benchmark for multi-label image classification in medical imaging, combining new classes and domains into a challenging scenario.

A variant of CIFAR-100 dataset, where the original dataset is split into 20 disjoint tasks, each consisting of 2,500 samples from 5 classes.

Generative kernel continual learning dataset

Continual Test Time Adaptation (CTTA) is re-required to adapt efficiently to continuous unseen domains while retaining previously learned knowledge.

The dataset used in the Continual Robot Imitation Learning (CRIL) paper, which consists of pseudo demonstrations of learned tasks and real demonstrations of new tasks.

The dataset used in the paper is a continual learning benchmark, consisting of three datasets: splitMNIST, splitFashionMNIST, and splitCIFAR.

The dataset used in the paper is not explicitly described, but it is mentioned that the authors used the CIFAR-100 and ImageNet-R benchmarks for class-incremental continual...

The Continual World benchmark consists of ten realistic robotic manipulation tasks.

A federated and continual learning framework for classification tasks in a society of devices

The paper proposes a Class-Prototype conditional Diffusion Model (CPDM) for continual learning with generative replay. The model uses a diffusion-based generator and a...

The dataset used in this paper is a custom dataset created for the CIFAR-100 class-incremental continual learning task.

The Sparse Diffusion Policy (SDP) framework, which integrates Mixture of Experts (MoE) layers into the diffusion policy.

FashionMNIST and MNIST datasets are used for continual unsupervised learning with variational autoencoders and generative replay.

The dataset used in the paper is the Split MS-COCO dataset, which is a comprehensive framework for continual image captioning.

Real-world time-series benchmark datasets for crude oil prices with distribution shifts

The dataset used in this paper for Continual Learning, Catastrophic Forgetting and PCA-OGD.

The Permuted MNIST dataset is a variation of MNIST where new tasks of comparable difficulty to the original MNIST classification task are created by permuting the pixels of...

The dataset used in this paper for Continual Learning, Catastrophic Forgetting and PCA-OGD.

The Rotated MNIST dataset is a subset of the MNIST dataset with images of handwritten digits rotated by 90 degrees.

Few-Shot Class Incremental Learning (FSCIL) is a challenging continual learning task, where limited training examples are available during several learning sessions.