79 datasets found

The dataset used in the paper for self-supervised learning of video representations.

Four datasets are relevant for our study: HMDB51, UCF101, Kinetics-400, and IG65M.

The DAVIS dataset comprises 60, 30 and 30 video sequences for training, validation and testing, respectively. The YouTube-VOS dataset is a larger dataset with 3471 videos for...

The Every Moment Counts dataset contains 1,000 hours of video footage of various activities.

The Hollywood in Homes dataset contains 9,848 videos of daily activities across 157 classes.

The THUMOS Challenge dataset contains 413 sports videos of 65 classes.

Temporal Action Detection (TAD) aims to identify the action boundaries and the corresponding category within untrimmed videos.

The Multi Visual Modality Fall Detection Dataset (MUVIM) was used for anomaly detection of falls. It contains (6) vision-based sensors of different modalities including thermal,...

UCF-101 dataset is a large-scale action recognition dataset, containing 13,320 videos categorized into 101 human action categories.

The Jester dataset is of continuous jokes ratings from -10 to 10, containing the jokes’ texts.

The DAVIS dataset is a widely used dataset for video-related tasks, consisting of approximately 2000 frames from 26 human-centric scenarios.

The MUSES dataset is a collection of 3,697 videos, with 2,587 for training and 1,110 for testing.

Temporal action localization (TAL) is a prevailing task due to its great application potential. Existing works in this field mainly suffer from two weaknesses: (1) They often...

Temporal action localization (TAL) is a challenging task in video understanding that aims to identify and localize actions within a video sequence.

Video classification is a fundamental problem in many video-based tasks. Applications such as autonomous driving technology, controlling drones and robots are driving the demand...

Mini-Kinetics-200: A dataset of 200 human action classes from videos in the wild.

Video recognition aims to automatically analyze the contents of videos (e.g., events and actions), and has a wide range of applications, including intelligent surveillance,...

The Kinetics-600 dataset consists of 392k training videos and 30k validation videos in 600 human action categories.

FineGym is a hierarchical video dataset for fine-grained action understanding, containing 354 action categories.

Motion has shown to be useful for video understanding, where motion is typically represented by optical flow. However, computing flow from video frames is very time-consuming....