Pretraining Image Encoders without Reconstruction via Feature Prediction Loss.
Gustav Grund PihlgrenFredrik SandinMarcus LiwickiPublished in: CoRR (2020)
Keyphrases
- image features
- high resolution
- image content
- reconstruction method
- input image
- image data
- single image
- multiscale
- image classification
- image segmentation
- image retrieval
- image analysis
- higher resolution
- segmentation method
- region of interest
- reconstruction process
- image representation
- prediction accuracy
- similarity measure
- invariant features
- reconstruction error
- computerized tomography
- spatial information
- three dimensional
- aerial images
- image processing
- compressed sensing
- scale space
- piecewise planar