* @brief * * @param frameA * @param frameB * @param matches * @return Eigen::Matrix4f : cam pose residual A-->B in world frame */

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

Args: resize (int, optional): the longer edge of resized images. None for no resize. padding (bool): If set to 'True', zero-pad r

Args: resize (tuple): align image to depthmap, in (w, h). augment_fn (callable, optional): augments images with pre-defined visua

compute a rotation exponential using the Rodrigues Formula. rotation axis w (theta = |w|); A = sin(theta) / theta; B = (1 - cos(theta)) / theta^2

@reader: data reader, sometimes we want to use the full resolution raw image

We first connect the discontinuous boxes for each ray and treat it as uniform sample, then we disconnect into correct boxes @rays_d: (N_ray,3)

Set global formatter. Each sink in each logger will get a clone of this object

Set global logging level

* @brief solve for pose 1 to 2 http://nghiaho.com/?page_id=671 * * @param points1 : Nx3 * @param points2 * @param pose */

@pts: shape can be (...,N,3 or 2) or (N,3) will homogeneliaze the last dimension

@coordinates: (3,N)

Return utc offset in minutes or throw spdlog_ex on failure

Warp kpts0 from I0 to I1 with depth, K and Rt Also check covisibility and depth consistency. Depth is consistent if relative error < 0.2 (har

Args: expec_f (torch.Tensor): [M, 2] <x, y> expec_f_gt (torch.Tensor): [M, 2] <x, y>

Args: expec_f (torch.Tensor): [M, 3] <x, y, std> expec_f_gt (torch.Tensor): [M, 2] <x, y>

Average Distance of Model Points for objects with no indistinguishable views - by Hinterstoisser et al. (ACCV 2012).

Add new frames and continue training @images: (N,H,W,3) new images @poses: All frames, they need to reset

@pred: 4x4 mat @gt: @model: (N,3)