Designing Rotationally Invariant Neural Networks from PDEs and Variational Methods.
Tobias AltKarl SchraderJoachim WeickertPascal PeterMatthias AugustinPublished in: CoRR (2021)
Keyphrases
- variational methods
- rotationally invariant
- neural network
- level set
- optic flow
- scale space
- curve evolution
- partial differential equations
- multiscale
- moment invariants
- approximate inference
- binary patterns
- pattern recognition
- optical flow estimation
- level set method
- optical flow
- active contours
- image segmentation
- higher order
- flow field
- shape prior
- coarse to fine
- object detection
- wavelet transform
- multiresolution
- image analysis
- feature space