Copied from Plenoxels Continuous learning rate decay function. Adapted from JaxNeRF The returned rate is lr_init when step=0 and lr_fina

Generic method iterating through a base_path folder to find scenes.

renorm pointmaps pts1, pts2 with norm_mode

Modify camera intrinsics to follow a different convention. Coordinates of the center of the top-left pixels are by default: - (0.5, 0.5)

tqdm but with parallel execution. Will essentially return res = [ function(arg) # default function(*arg) # if star_args is

Modification based on TORCHVISION.UTILS ::param: tensor (batch, channel, H, W)

Predict in the correct aspect-ratio, then transpose the result in landscape and stack everything back together.

visualize predicted pose result vs. gt pose

Init. Args: scale_factor (float): scaling mode (str): interpolation mode

Backprojection operation: from image depths to 3D points

Method: 1) express all 3d points in each camera coordinate frame 2) if they're in front of a depthmap --> then lower their confidence

Reprojection method, for when the absolute depth is known: 1) estimate the camera focal using a robust estimator 2) reproject points

set batchnorm to eval() it is useful to align train and testing result

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

extract confidence from prediction head output

flow_utils._computeColor(flow): compute color codes for the flow field flow ---- PARAMETERS ---- flow: np.array of dimension (he

Loads the full state from disk if it exists. Returns the next epoch to continue training from.