Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learning.
Colorado J. ReedRitwik GuptaShufan LiSarah BrockmanChristopher FunkBrian ClippKurt KeutzerSalvatore CandidoMatt UyttendaeleTrevor DarrellPublished in: CoRR (2022)
Keyphrases
- multiscale
- learning process
- scale space
- learning algorithm
- scale selection
- learning systems
- image representation
- unsupervised learning
- prior knowledge
- background knowledge
- learning tasks
- multiple scales
- active learning
- online learning
- neural network
- spatial data
- learning problems
- learning analytics
- feature space
- single scale