processes the 1x12 raw pose from dataset by aligning and then normalizing produce logq :param poses_in: N x 12 :return: processed poses N x 1

Applies logarithm map to q :param q: (4,) :return: (3,)

render nerfw imgs, save unscaled pose and results

Compute RGB MSE Loss, original from NeRF Paper

Prepares inputs and applies network 'fn'.

We need a new query function, Coarse = NeRF, Fine = NeRF-W Inputs: inputs: torch.Tensor() [N_rays,N_samples,3] viewdirs: torch.T

Perform inference on a random val image and save the result

core training loop for featurenet

Perform 1 step of training

Perform 1 step of training

we implement random view synthesis for generating more views to help training posenet

Perform 1 epoch of training with batch

Perform 1 epoch of training with batch

Warp a list of points with the given homography. Arguments: points: list of N points, shape (N, 2(x, y))). homography: batch

In: observation: Nx3 Out: pred_state: Nx3

Args: tensor (Tensor): Tensor image of size (C, H, W) to be normalized. Returns: Tensor: Normalized image.

Args: patience (int): How long to wait after last time validation loss improved. Default: 50

lambda_u: in equation 13

---Parameters for the original NeRF--- D: number of layers for density (sigma) encoder W: number of hidden units in each laye

Initialize one adaptation layer for every extraction point. Args: hypercolumn_layers: The list of the hypercolumn layer names.

Initialize one adaptation layer for every extraction point. Args: hypercolumn_layers: The list of the hypercolumn layer names.

:param scene: scene name ['chess', 'pumpkin', ...] :param data_path: root 7scenes data directory. Usually '../data/deepslam_data/7S

:param scene: scene name ['chess', 'pumpkin', ...] :param data_path: root 7scenes data directory. Usually '../data/deepslam_data/7S

assume target and nerf rgb are inferenced at the same time, slice target batch and nerf batch and aggregate features :param x

Compute a boolean mask of the valid pixels resulting from an homography applied to an image of a given shape. Pixels that are False correspon

downscale translation pose to [-1:1] only

local NeRF data selection

:param x: image blob (N x T x C x H x W) :return: pose outputs (N x T x 6)

Encodes input (xyz+dir) to rgb+sigma (not ready to render yet). For rendering this ray, please see rendering.py Inputs:

Apply adaptation layers. # here is list of three levels of features

inference DFNet. It can regress camera pose as well as extract intermediate layer features. :param x: image blob (2B x C x H x W)

inference DFNet_s. It can regress camera pose as well as extract intermediate layer features. :param x: image blob (2B x C x H x

Apply adaptation layers. # here is list of three levels of features

inference DFNet. It can regress camera pose as well as extract intermediate layer features. :param x: image blob (2B x C x H x W)

Apply adaptation layers. # here is list of three levels of features

inference DFNet. It can regress camera pose as well as extract intermediate layer features. :param x: image blob (2B x C x H x W)