HACS

Human Action Clips and Segments

Dataset Information
Modalities
Videos
License
BSD 3-Clause
Homepage
Official Website
Contents

Overview

HACS is a dataset for human action recognition. It uses a taxonomy of 200 action classes, which is identical to that of the ActivityNet-v1.3 dataset. It has 504K videos retrieved from YouTube. Each one is strictly shorter than 4 minutes, and the average length is 2.6 minutes. A total of 1.5M clips of 2-second duration are sparsely sampled by methods based on both uniform randomness and consensus/disagreement of image classifiers. 0.6M and 0.9M clips are annotated as positive and negative samples, respectively.

Authors split the collection into training, validation and testing sets of size 1.4M, 50K and 50K clips, which are sampled
from 492K, 6K and 6K videos, respectively.

Variants: HACS

Associated Benchmarks

This dataset is used in 2 benchmarks:

Recent Benchmark Submissions

Task Model Paper Date
Temporal Action Localization RDFA-S6 (InternVideo2-6B) Enhancing Temporal Action Localization: Advanced … 2024-07-18
Temporal Action Localization DyFADet(VideoMAEv2) DyFADet: Dynamic Feature Aggregation for … 2024-07-03
Action Recognition InternVideo2-6B InternVideo2: Scaling Foundation Models for … 2024-03-22
Temporal Action Localization InternVideo2-6B InternVideo2: Scaling Foundation Models for … 2024-03-22
Temporal Action Localization InternVideo2-1B InternVideo2: Scaling Foundation Models for … 2024-03-22
Temporal Action Localization ActionMamba(InternVideo2-6B) Video Mamba Suite: State Space … 2024-03-14
Temporal Action Localization TriDet (VideoMAEv2) Temporal Action Localization with Enhanced … 2023-09-11
Temporal Action Localization TriDet (SlowFast) TriDet: Temporal Action Detection with … 2023-03-13
Temporal Action Localization TriDet (I3D RGB) TriDet: Temporal Action Detection with … 2023-03-13
Temporal Action Localization InternVideo InternVideo: General Video Foundation Models … 2022-12-06
Action Recognition UniFormerV2-L UniFormerV2: Spatiotemporal Learning by Arming … 2022-09-22
Temporal Action Localization TadTr (I3D RGB) End-to-end Temporal Action Detection with … 2021-06-18
Action Recognition SRTG r3d-101 Learn to cycle: Time-consistent feature … 2020-06-15
Action Recognition SRTG r(2+1)d-101 Learn to cycle: Time-consistent feature … 2020-06-15
Action Recognition SRTG r(2+1)d-50 Learn to cycle: Time-consistent feature … 2020-06-15
Action Recognition SRTG r3d-34 Learn to cycle: Time-consistent feature … 2020-06-15
Action Recognition SRTG r(2+1)d-34 Learn to cycle: Time-consistent feature … 2020-06-15
Action Recognition SRTG r3d-50 Learn to cycle: Time-consistent feature … 2020-06-15
Temporal Action Localization SSN HACS: Human Action Clips and … 2017-12-26

Research Papers

Recent papers with results on this dataset:

External Links: