point_samples [N_rays N_sample 3]

computes the percentage of pixels whose depth error is less than @threshold

depth: (H, W)

sample points along the rays Inputs: rays: () Returns:

Get ray directions for all pixels in camera coordinate. Reference: https://www.scratchapixel.com/lessons/3d-basic-rendering/ r

Get ray origin and normalized directions in world coordinate for all pixels in one image. Reference: https://www.scratchapixel.com/lessons/3d

Normalize a vector.

Instantiate mvs NeRF's MLP model.

src_feat: (B, C, H, W) proj_mat: (B, 3, 4) equal to "src_proj @ ref_proj_inv" depth_values: (B, D, H, W) out: (B, C, D, H, W)

Args: volume_color_feature: [B, G, D, h, w] volume_density_feature: [B C D H W] ray_dir_world:[3 ray_samp

point_world: [N_ray N_sample 3] imgs: [N V 3 H W]

Center the poses so that we can use NDC. See https://github.com/bmild/nerf/issues/34 Inputs: poses: (N_images, 3, 4) Outputs:

Normalize a vector.

Calculate the average pose, which is then used to center all poses using @center_poses. Its computation is as follows: 1. Compute the cen

Constructs a version of 'fn' that applies to smaller batches.

Args: imgs: [N V C H W] depths: [N V H W] poses: w2c c2w intrinsic [N V 4 4] [B V levels 3 3)] init_depth_min: [

Args: extrinsic intrinsic [4 4] [3 3)] N_samples: same as D int depth_values: [B D] Returns: [3 N_rays N_sa

Inputs: c2ws: [1,v,4,4] rays_pts: [N_rays, N_samples, 3] rays_dir: [N_rays, 3] Returns:

Transform rays from world coordinate to NDC. NDC: Space such that the canvas is a cube with sides [-1, 1] in each axis. For detailed deri

Transforms model's predictions to semantically meaningful values. Args: raw: [num_rays, num_samples along ray, 4]. Prediction from model.

Prepares inputs and applies network 'fn'.

canonical_volume, canonical_density, canonical_weightsl: [1,C,D,H,W] ray_feat, ray_ndc_pts, ray_weight: [N_ray, N_sample, C]