A Short Note on the Kinetics-700-2020 Human Action Dataset
By Lucas Smaira, João Carreira, Eric Noland, Ellen Clancy, Amy Wu, Andrew Zisserman. 2020
173 citations recorded by the source corpus.
We describe the 2020 edition of the DeepMind Kinetics human action dataset, which replenishes and extends the Kinetics-700 dataset. In this new version, there are at least 700 video clips from different YouTube videos for each of the 700 classes. This paper details the changes introduced for this new release of the dataset and includes a comprehensive set of statistics as well as baseline results using the I3D network.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Evaluation & Benchmarks
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Quoc V. Le
- Andrew Zisserman
- Li Fei-Fei
- Sergey Levine
- Trevor Darrell
- Yoshua Bengio
- Luc Van Gool
- Mohit Bansal
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Andrew Zisserman
- Dacheng Tao
- Jitendra Malik
- Trevor Darrell
- Li Fei-Fei
- Cordelia Schmid
- Ming-Hsuan Yang
In Evaluation & Benchmarks
- Andrew Zisserman
- João Carreira
- Li Fei-Fei
- Jiahui Yu
- Jiajun Wu
- Jean-Baptiste Alayrac
- Dima Damen
- Juan Carlos Niebles
Citations in the corpus
- Perceiver IO: A General Architecture for Structured Inputs & Outputs (2021, methodology): “These are then linearly projected to the appropriate channel size for each modality: 3 for videos, 16 for audio and 700 for classification (the logits for each of the 700 classes in Kinetics700).”
- Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language (2022, background): “These activity suggestions are often more expressive than the class categories that can be found in typical activity recognition datasets (e.g., Charades (Sigurdsson et al. 2018), Kinetics (Smaira et al. 2020)), and reflect the information already stored in the model, agnostic to the point of view.”
- CoCa: Contrastive Captioners are Image-Text Foundation Models (2022, methodology): “Many of the videos have been made explicitly unavailable [72], hence we compute retrieval over the subset of data that is publicly available at the time of evaluation.”
- Diffusion Probabilistic Modeling for Video Generation (2022, methodology): “Cityscape dataset only contains traffic video recorded by a camera installed in the front of the car), as training a large model for multi-domain datasets like Kinetics (Smaira et al., 2020) is demanding for our”
- Revisiting the “Video” in Video-Language Understanding (2022, background): “The impact of such analysis was tremendous: later datasets were designed to capture a richer distribution of temporal understanding [6,13,47] with better dis-entanglement of such cues [34], and model designs evolved further to better capture the now necessary dynamics to address these improved…”
- Learning from One Continuous Video Stream (2023, methodology): “We pre-trained on the Kinetics-700-2020 dataset [40], which is composed of 10s clips.”