:param pts: (N, 3 or 2) :return pts_hom: (N, 4 or 3)

Copied from Detectron codebase.

extract features of pts for rendering :param pts: :param geometryVolume: :param vol_dims: :param partial_vol_

:param xyz: [N_rays, n_samples,3 ] :param query_c2w: [1,4,4] :param supporting_c2ws: [n,4,4] :return:

:param xyz: [N_rays, n_samples,3 ] :param surface_normals: [N_rays, n_samples,3 ] :param supporting_c2ws: [n,4,4] :r

Compute the max candidates of all pairs within a batch Args: p_m0, p_m1 (torch.Tensor): padded masks

extract features of pts for rendering :param pts: :param geometryVolume: :param vol_dims: :param partial_vol_

if use ddp, use setup() not prepare_data(), prepare_data() only called on 1 GPU/TPU in distributed :return:

if use ddp, use setup() not prepare_data(), prepare_data() only called on 1 GPU/TPU in distributed :return:

Use the pred_depthmaps to remove redundant pts (pruned by sdf, sdf always have two sides, the back side is useless) :param coords: [n

Args: conf_matrix (torch.Tensor): [N, L, S] data (dict): with keys ['hw0_i', 'hw1_i', 'hw0_c', 'hw1_c'] Retur

This matches the implementation in Denoising Diffusion Probabilistic Models: From Fairseq. Build sinusoidal embeddings. This matches

Get corners of 3D camera view frustum of depth image

A util function for plotting a grid of images. Args: images: (N, H, W, 4) array of RGBA images rows: number of rows in the gr

imgs: list, N images of size WxHxC