66 datasets found

Action recognition is a fundamental topic in video understanding and has made significant progress in recent years.

A video dataset for action recognition, consisting of 920 videos of 21 different actions.

The H36M-4A dataset is derived from the H36M-Original dataset, using the 4A pipeline to generate synthetic data.

The H36M-Original dataset consists of 6 action classes, with 16 frames per class.

The NTU-4A dataset is derived from the NTU-Original dataset, using the 4A pipeline to generate synthetic data.

The NTU-Original dataset consists of 49 single-person action classes from the real world.

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

The HumanEva-I dataset is a dataset for 3D human pose estimation, containing video sequences of four subjects performing six common actions.

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

Resound is a video dataset for action recognition without representation bias.

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....

The Something-Something V1 & V2 dataset is a large-scale video dataset created by crowdsourcing. It contains about 100k videos over 174 categories, and the number of videos...

UIUC-Sports dataset

The UCF101 dataset contains 13320 videos distributed in 101 action categories. This dataset is different from the above ones in that it contains mostly coarse sports activities...

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...

The dataset contains 3D human skeletons and corresponding action labels.

The dataset contains 3D human skeletons and corresponding action labels.

The MSR-Action3D dataset has been first introduced with a depth-based approach, but its joint streams were widely used later as a reference for joint-based methods.

Kinetics-700 is a large-scale video dataset for human action recognition, with 700 action categories.

The dataset used for video action classification, consisting of 9.8k training videos, 1.8k validation videos, and 157 classes.