Feature extraction Module that extracts hybrid features

Multi-layer perception. [B, Cin, N] -> [B, Cout, N] or [B, Cin] -> [B, Cout]

Generic N layer decoder with masking.

Decoder is made of self-attn, src-attn, and feed forward (defined below)

A standard Encoder-Decoder architecture. Base for this and many other models.

Implements FFN equation.

Early fusion implementation of RPMNet, as described in the paper

Early fusion implementation of RPMNet, as described in the paper

Early fusion implementation of RPMNet, as described in the paper

Manager for saving/managing pytorch checkpoints. Provides functionality similar to tf.Saver such as max_to_keep and keep_checkpoint_every_n_h

Exp: se(3) -> SE(3)

Exp: so(3) -> SO(3)

Exp: se(3) -> SE(3)

Exp: so(3) -> SO(3)

M(n) -> M(n); x -> x^-1.

generate perturbations

rigid motion

[N, D] -> [M, D]

Enters debugger when anomaly detected