static Acc_from_confusions(confusions)#
static IoU_from_confusions(confusions)#

Computes IoU from confusion matrices.


confusions – ([…, n_c, n_c] np.int32). Can be any dimension, the confusion matrices should be described by

the last axes. n_c = number of classes


([…, n_c] np.float32) IoU score

static cam2img(points, cam_img)#
static cam2world(points, world_cam)#
static data_aug(xyz, color, labels, idx, num_out)#
static get_class_weights(num_per_class)#
static grid_subsampling(points, features=None, labels=None, grid_size=0.1, verbose=0)#

CPP wrapper for a grid subsampling (method = barycenter for points and features).

  • points – (N, 3) matrix of input points

  • features – optional (N, d) matrix of features (floating number)

  • labels – optional (N,) matrix of integer labels

  • grid_size – parameter defining the size of grid voxels

  • verbose – 1 to display


Subsampled points, with features and/or labels depending of the input

static invT(T)#

KNN search.

  • support_pts – points you have, N1*3

  • query_pts – points you want to know the neighbour index, N2*3

  • k – Number of neighbours in knn search


neighboring points indexes, N2*k

Return type:


static load_label_kitti(label_path, remap_lut)#
static load_label_semantic3d(filename)#
static load_pc_kitti(pc_path)#
static load_pc_semantic3d(filename)#
static remove_outside_points(points, world_cam, cam_img, image_shape)#

Remove points which are outside of image.

  • points (np.ndarray, shape=[N, 3+dims]) – Total points.

  • world_cam (np.ndarray, shape=[4, 4]) – Matrix to project points in lidar coordinates to camera coordinates.

  • cam_img (p.array, shape=[4, 4]) – Matrix to project points in camera coordinates to image coordinates.

  • image_shape (list[int]) – Shape of image.


Filtered points.

Return type:

np.ndarray, shape=[N, 3+dims]

static shuffle_idx(x)#
static shuffle_list(data_list)#
static world2cam(points, world_cam)#