VideoMAE

GPTKB entity

Statements (23)
Predicate Object
gptkbp:instanceOf video understanding model
gptkbp:application action recognition
video classification
video representation learning
gptkbp:architecture gptkb:Vision_Transformer_(ViT)
gptkbp:author Zhan Tong
gptkbp:basedOn Masked Autoencoders (MAE)
gptkbp:citation VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
gptkbp:developedBy gptkb:Microsoft_Research_Asia
https://www.w3.org/2000/01/rdf-schema#label VideoMAE
gptkbp:improves data efficiency in video pretraining
gptkbp:input gptkb:DVD
gptkbp:language gptkb:Python
gptkbp:notableFeature end-to-end transformer architecture
high masking ratio for video frames
gptkbp:openSource yes
gptkbp:pretrainingMethod masked video modeling
gptkbp:publicationYear 2022
gptkbp:publishedIn ECCV 2022
gptkbp:repository https://github.com/MCG-NJU/VideoMAE
gptkbp:uses self-supervised learning
gptkbp:bfsParent gptkb:Hugging_Face_models
gptkbp:bfsLayer 7