The Kinetics Human Action Video Dataset
This dataset provides a standardized benchmark for training and evaluating video-based human action recognition models, benefiting researchers in computer vision and machine learning.
The authors introduced the Kinetics dataset, a large-scale video dataset with 400 human action classes and at least 400 clips per class, totaling over 160,000 clips, to address the need for diverse and extensive data in human action recognition, and they provided baseline performance figures for neural networks trained on it.
We describe the DeepMind Kinetics human action video dataset. The dataset contains 400 human action classes, with at least 400 video clips for each action. Each clip lasts around 10s and is taken from a different YouTube video. The actions are human focussed and cover a broad range of classes including human-object interactions such as playing instruments, as well as human-human interactions such as shaking hands. We describe the statistics of the dataset, how it was collected, and give some baseline performance figures for neural network architectures trained and tested for human action classification on this dataset. We also carry out a preliminary analysis of whether imbalance in the dataset leads to bias in the classifiers.