revert image_hwc_to_chw function

convert traj_str into translation and rotation matrices Args: traj_str: A space-delimited file where each line represents a camera positi

7 -------- 4 /| /| 6 -------- 5 . | | | | . 3 -------- 0 |/ |/ 2 -------- 1 A

extract original label data Args: gt_fn: str (file name of "annotation.json") after loading, we got a dict with keys

transpose the image from height, width, channel -> channel, height, width (pytorch format)

corners: (8,3) no assumption on axis direction

Args: pose: np.array (4, 4) Returns: index: int (0, 1, 2, 3) for upright, left, upside-down and right

Generic functions to compute precision/recall for object detection for a single class. Args: pred: map of {img_id: [(bbox, score)]} wh

Args: box: 8x3 Returns: size: [dx, dy, dz]

Ref: http://stackoverflow.com/questions/24467972/calculate-area-of-polygon-given-x-y-coordinates

:param ply_file: path of ply file :param json_file: path of annotation result json file :param back_face_cull: see single sid

Args: loader: TenFpsDataLoader Returns: world_pc: (N, 3) xyz in world coordinate system world_sem: (N, d)

In the backward pass we receive a Tensor containing the gradient of the loss with respect to the output, and we need to compute the g

Compute corners of a single box from rotation matrix Args: size: list of float [dx, dy, dz] center: np.array [x, y, z] rot

Return a Matrix3 for the angle axis. Arguments: angle_axis {Point3} -- a rotation in angle axis form.

Compute area of two convex hull's intersection area. p1,p2 are a list of (x,y) tuples of hull vertices. return a list of (x,y) for th

Euler rotation matrix with clockwise logic. Rotation Args: theta: list of float [theta_x, theta_y, theta_z] Returns:

expand image dimension to add a channel dimension

Generate 3D point coordinates and related rgb feature Args: rgb_image: (h, w, 3) rgb depth_image: (h, w) depth intrinsic:

Args: box: (8, 3) Returns: heading_angle: float

Args: box: np.array (8, 3) 8 corners color: line color Returns: o3d.Linset()

Args: points: (N, 3) boxes: (m, 8, 3) Returns: votes: (N, 4)

Args: pc: np.array (n, 3+d) boxes: np.array (m, 8, 3) Returns: mask: np.array (n, m) of type bool

Clip a polygon with another polygon. Ref: https://rosettacode.org/wiki/Sutherland-Hodgman_polygon_clipping#Python Args: subjectPolygon:

Rotation clockwise Args: points: np.array of np.array (B, N, 3 + C) or (N, 3 + C) for single batch angle: np.array of

Accumulate one batch of prediction and groundtruth. self.gt_map_cls[idx]: a list of (int, np.array (8, 3)) self.gt_map_cls[idx

Visualize result with open3d Args: pc: np.array of shape (n, 3) point cloud boxes: a list of m boxes, each item a

ap = voc_ap(rec, prec, [use_07_metric]) Compute VOC AP given precision and recall. If use_07_metric is true, uses the VOC 07 11 point meth

Args: index (int): Index Returns: tuple: (identifier, color, highres_depth, lowres_depth).

Returns: frame: a dict {frame_id}: str {depth}: (h, w) {image}: (h, w)