Main function to train the model

Pairwise nearest neighbors for L2-normalized vectors. Uses Torch rather than Faiss to remain on GPU.

x: (N, L, patch_size**2 *3) imgs: (N, 3, H, W)

imgs: (N, 3, H, W) x: (N, L, patch_size**2 *3)

imgs: (N, 3, T, H, W) x: (N, L, patch_size**2 *3*T)

imgs: (N, 3, H, W) x: (N, L, patch_size**2 *3)

Prints content of DictConfig using Rich library and its tree structure. Args: cfg (DictConfig): Configuration composed by Hydra.

Given a quaternion representing rotation, get the quaternion representing its inverse. Args: quaternion: Quaternions as tensor o

Constructs a ResNet model. Args: num_layers (int): Number of resnet layers. Must be 18 or 50 pretrained (bool): If True, returns a

Convert an axisangle rotation into a 4x4 transformation matrix (adapted from https://github.com/Wallacoloo/printipi) Input 'vec' has to be Bx1

The First order smoothness loss. Modified from https://github.com/lliuz/ARFlow/blob/master/losses/flow_loss.py licensed under MIT License

The training loop running a single training epoch. Args: module: the LightningModule to train optimizer: the optimize

Convert the network's (axisangle, translation) output into a 4x4 matrix

Upsample input tensor by a factor of 2

The validation loop ruunning a single validation epoch. Args: module: the LightningModule to evaluate val_loader: The

warp an image/tensor (im2) back to im1, according to the optical flow x: [B, C, H, W] (im2) flo: [B, 2, H, W] flow

Create encoder and predictor masks when collating imgs into a batch # 1. sample pred block size using seed # 2. sample severa

Returns a single training item from the dataset as a dictionary.

Returns a single training item from the dataset as a dictionary.