40 datasets found

A new approach to meta-learning, called meta-meta classification, to learning in small-data settings.

Few-shot segmentation remains challenging due to the limitations of its labeling information for unseen classes. Most previous approaches rely on extracting high-level feature...

The proposed MG-ViT model is used for few-shot learning on the Agri-ImageNet and ACFR apple detection tasks.

The MiniVSPW dataset is a benchmark for few-shot video object segmentation.

The FewSol dataset is used for training the class-adaptive object detector. It contains 666 objects with a large number of occurrences.

A dataset for training and testing semantically consistent few-shot view synthesis models

The CIFAR-FS dataset is a few-shot learning dataset, consisting of 60,000 color images from 100 classes with 600 examples per class.

The ImageCLEF-DA dataset is a benchmark dataset for ImageCLEF 2014 domain adaptation challenges, which contains 12 categories shared by three domains: Caltech-256 (C), ImageNet...

A dataset of datasets for learning to learn from few examples.

MiniImageNet, Caltech-UCSD Birds 200-2011, TieredImageNet, OfficeHome

Few-shot semantic segmentation via prototype augmentation with image-level annotations

This dataset is used for few-shot novel view synthesis.

The MiniImagenet dataset is a benchmark for few-shot learning, consisting of 60,000 images from 21 classes, each with 300 images.

PartImageNet and Pascal Part datasets for few-shot part segmentation.

Few-shot part segmentation using few-shot support images and pre-trained image-language model CLIP.

Gaussian processes (GPs) are non-parametric, flexible models that work well in many tasks. Combining GPs with deep learning methods via deep kernel learning (DKL) is especially...

Zero-shot learning (ZSL) aims to classify objects that are not observed or seen during training. It relies on class semantic description to transfer knowledge from the seen...

Few-Shot Class-Incremental Learning (FSCIL) is a special case of Class-Incremental Learning (CIL), where only a few training examples are available at every learning session.

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

The Human Connectome Project (HCP) dataset contains volumetric task fMRI activation maps from the Human Connectome Project 1200 dataset (HCP1200) distribution, for the 965...