Model and their names

Models

Model	Name in Code	Bases	Paper
RST-CNN	mnist_ses_scalar_28_rot_8	E	Deformation Robust Roto-Scale-Translation Equivariant CNNs
RST-CNN	mnist_ses_scalar_28_rot_8	G	Deformation Robust Roto-Scale-Translation Equivariant CNNs
RST-CNN+	mnist_ses_scalar_28_rot_8_interrot_4	E	Deformation Robust Roto-Scale-Translation Equivariant CNNs
RST-CNN+	mnist_ses_scalar_28_rot_8_interrot_4	G	Deformation Robust Roto-Scale-Translation Equivariant CNNs
SFCNN	mnist_res_scalar_28_rot_8	I	Learning Steerable Filters for Rotation Equivariant CNNs. ​
SFCNN+	mnist_res_scalar_28_rot_8_interrot_4	I	Learning Steerable Filters for Rotation Equivariant CNNs. ​
RDCF	mnist_res_scalar_28_rot_8	D1	RotDCF: Decomposition of Convolutional Filters for Rotation-Equivariant Deep Networks
RDCF+	mnist_res_scalar_28_rot_8_interrot_4	D1	RotDCF: Decomposition of Convolutional Filters for Rotation-Equivariant Deep Networks​
SEVF	mnist_sevf_scalar_28	\	Scale equivariance in CNNs with vector fields​
SESN	mnist_ses_scalar_28_rot_1	C	Scale-Equivariant Steerable Networks
SDCF	mnist_ses_scalar_28_rot_1	G	Scale-Equivariant Neural Networks with Decomposed Convolutional Filters

Note here $+$ denote inter-rotation version of model implementation. Please specify model and bases in corresponding shell file.

Basis

• Basis A: Hermite Gaussian one scale
• Basis B: Hermite Gaussian multi scale
• Basis C: Hermite Gaussian rotation and multi-scale
• Basis D: Fourier Bessel one scale
• Basis E: Fourier Bessel rotation and multi-scale
• Basis F: Fourier Bessel Gaussian rotation and multi-scale
• Basis G: SL rotation and multi-scale