45 datasets found

The AMI Meeting Corpus is a multimodal corpus containing audio and video recordings of meetings.

The EasyCom dataset is a relatively new dataset, recorded using Meta’s Augmented-Reality (AR) glasses set.

Continuous sign language recognition (SLR) deals with unaligned video-text pair and uses the word error rate (WER), i.e., edit distance, as the main evaluation metric.

The ShapeNeRF–Text dataset consists of 40K paired NeRFs and language annotations for ShapeNet objects.

The MSVD-QA dataset is a benchmark for video question answering, containing 1,970 videos with multiple-choice questions.

The TGIF-QA dataset consists of 165165 QA pairs chosen from 71741 animated GIFs. To evaluate the spatiotemporal reasoning ability at the video level, TGIF-QA dataset designs...

The BraTS2020 dataset is a widely-used benchmark for brain tumor segmentation. It contains 369 glioma patient samples with two glioma grades (LGG and HGG).

The BraTS2018 database is a continually evolving database with a total of 285 glioblastoma or low-grade gliomas subjects, comprising three consecutive subsets, i.e., 30 subjects...

The dataset used in the paper for training and evaluation of the MERU method.

The dataset used in the paper for training and evaluation of the HYPE method.

Automatic evaluations for natural language generation (NLG) conventionally rely on token-level or embedding-level comparisons with the text references. This is different from...

A multimodal variational autoencoder for low-cost cardiac hemodynamics instability detection from CXR and ECG.

The CSL dataset is a large-scale Chinese scientific literature dataset obtained from the "Qianyan" open-source NLP platform. It consists of 396,209 Chinese core journal papers'...

PMC-CLIP: Contrastive language-image pre-training using biomedical documents.

BioMedClip: A CLIP model pretrained on image-text pairs extracted from PubMed Central repository.

Contrastive Language-Image Pretraining (CLIP) models are trained with datasets extracted from web crawls, which are of large quantity but limited quality. This paper explores...

The dataset used in the paper "Scaling Laws of Synthetic Images for Model Training". The dataset is used for supervised image classification and zero-shot classification tasks.

Few-shot image classification is a research area that focuses on identifying new classes with a small number of samples.

Few-shot action recognition is a promising direction to alleviate the data labeling problem, which aims to identify unseen classes with a few labeled videos.

Conceptual 12m dataset for automatic image captioning