Runs the secant method for interval [z_low, z_high] for n_secant_steps

Run sphere tracing algorithm for max iterations from both sides of unit sphere intersection

Compute voting of expected rotation and translation for each ray. Param: rays_o: input origins [N_imgs, N_rays, 3] ra

Nerf MLP backbone

Init layered sampling num_cams: number of training cameras trainable: Whether planes can be trained by optimizer

Init layered sampling init_planes: [N_planes, 4], Ax + By + Cz = D trainable: Whether planes can be trained by optimizer

Init layered sampling init_planes: [N_planes, 4], Ax + By + Cz = D trainable: Whether planes can be trained by optimizer

Single forward feed that applies to smaller batches.

Transforms model's predictions to semantically meaningful values. Args: raw: [num_rays, num_samples, C]. Prediction from model.

Transforms model's predictions to semantically meaningful values. Args: raw: [num_rays, num_samples along ray, C]. Prediction from

Generate sample points Args: rays_o: [N_rays, 3+1] origin points of rays with camera id rays_d: [N_rays, 3+1] directions of r

Render rays Args: ray_batch: array of shape [2, batch_size, 3]. Ray origin and direction for each example in batch.

Generate sample points Args: rays_o: [N_rays, 3] origin points of rays rays_d: [N_rays, 3] directions of rays

Generate sample points Args: rays_o: [N_rays, 3] origin points of rays rays_d: [N_rays, 3] directions of rays

Generate sample points Args: rays_o: [N_rays, 3] origin points of rays rays_d: [N_rays, 3] directions of rays render

Voting for the expected rotation and translation. Param: rays_o: input origins [N_imgs, N_rays, 3] rays_d: input dire

Prepares inputs and applies network. inputs: shape:[1024, 64, 3] viewdirs: shape:[1024, 3]

Retruns depth from 3D points according to camera pose