Something-Something

The Something-Something dataset (version 2) is a collection of 220,847 labeled video clips of humans performing pre-defined, basic actions with everyday objects. It is designed to train machine learning models in fine-grained understanding of human hand gestures like putting something into something, turning something upside down and covering something with something.

Papers using Something-Something (9)

Few-shot Video Classification Via Temporal Alignment2019 · 263 citesTemporal-relational Crosstransformers For Few-shot Action Recognition2021 · 187 citesCo-training Transformer With Videos And Images Improves Action Recognition2021 · 31 citesMotion-guided Masking For Spatiotemporal Representation Learning2023 · 17 citesMotion Guided Attention Fusion To Recognize Interactions From Videos2021 · 16 citesOn The Surprising Effectiveness Of Transformers In Low-labeled Video Recognition2022 · 1 citesInsights from Visual Cognition: Understanding Human Action Dynamics with Overall Glance and Refined Gaze Transformer2026SVT: Supertoken Video Transformer For Efficient Video Understanding2023ViViT: A Video Vision Transformer2021