Read all the lines in a text file and return as a list

Run the entire training pipeline

Write an event to the tensorboard events file

Convert time in seconds to a nice string e.g. 10239 -> '02h50m39s'

Run the entire training pipeline

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

Convert network's sigmoid output into depth prediction The formula for this conversion is given in the 'additional considerations' section of

Run the entire training pipeline

Upsample input tensor by a factor of 2

Computes the smoothness loss for a disparity image The color image is used for edge-aware smoothness

calculate the normalize local cross correlation between I and J assumes I, J are sized [batch_size, *vol_shape, nb_feats]

Generate the warped (reprojected) color images for a minibatch. Generated images are saved into the `outputs` dictionary.

Generate the warped (reprojected) color images for a minibatch. Generated images are saved into the `outputs` dictionary.

Computes the smoothness loss for a appearance flow

Write an event to the tensorboard events file

Write an event to the tensorboard events file

Rescale image pixels to span range [0, 1]

Predict poses between input frames for monocular sequences.

Predict poses between input frames for monocular sequences.

Convert all models to training mode

Convert all models to training mode

Convert all models to training mode

Apply the disparity post-processing method as introduced in Monodepthv1

Computation of error metrics between predicted and ground truth depths

Compute the reprojection, perception_loss and smoothness losses for a minibatch

Compute the reprojection, perception_loss and smoothness losses for a minibatch

Compute the reprojection, perception_loss and smoothness losses for a minibatch

Evaluate odometry on the SCARED dataset

Evaluates a pretrained model using a specified test set

:param data: unnormalized coordinates Bx2xHxW :return: Bx1xHxW

Computes the smoothness loss for a optical flow

Convert a translation vector into a 4x4 transformation matrix

Load model(s) from disk

Load model(s) from disk

Load model(s) from disk

Print a logging statement to the terminal

Print a logging statement to the terminal

Print a logging statement to the terminal

Predict poses between input frames for monocular sequences.

Resize colour images to the required scales and augment if required We create the color_aug object in advance and apply the same augmentation

Pass a minibatch through the network and generate images and losses

Pass a minibatch through the network and generate images and losses

Pass a minibatch through the network and generate images and losses

Pass a minibatch through the network and generate images and losses

Pass a minibatch through the network and generate images and losses

Pass a minibatch through the network and generate images and losses

Pass a minibatch through the network and generate images and losses

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

Run a single epoch of training and validation

Run a single epoch of training and validation

Run a single epoch of training and validation

Save model weights to disk

Save model weights to disk

Save model weights to disk

Save options to disk so we know what we ran this experiment with

Save options to disk so we know what we ran this experiment with

Save options to disk so we know what we ran this experiment with

Convert time in seconds to time in hours, minutes and seconds e.g. 10239 -> (2, 50, 39)

Convert all models to testing/evaluation mode

Convert all models to testing/evaluation mode

Convert all models to testing/evaluation mode

Convert all models to training mode

Function to predict for a single image or folder of images

Validate the model on a single minibatch

Validate the model on a single minibatch

Returns a single training item from the dataset as a dictionary. Values correspond to torch tensors. Keys in the dictionary are eithe