Generates all 3 representations of floor plans from data in boxes npz and write them to boxes npz.

Run test pass on the dataset

Adapted from tablechair_horizontal_success, used to report numerical results for further analysis.

Input: x - B, N, 3 Output: output_dictionary - dictionary with all outputs and inputs

Spherical Harmonics Transform to extract type 0 features

Spherical Harmonics Transform to extract type 1 features that are equivariant to SO(3)

Graph a set of points. xy: numpy array of shape (numpt,2), positions of points to visualize

Graph 2 sets of corresponding points, with arrows from xy_1 to xy_2. xy_1: has shape (numpt,>=2), ex: input position to network xy_2:

width and height: res*2. Top left is (0,0) x: has shape (6,2)

Example: x = generate_6_points_batch(batch_size=batch_size)

Old/simple way of 3d visualization. final_pos, final_angle, initial_sha: [14, 2], assume first 2 rows are tables, and the 2:numobj are chair

Graph 2 sets of tables and chairs, with arrows from xy_1 to xy_2. xy_1: has shape (numpt,6), ex: input position to network xy_2: has

posang has shape (1,14,>=2). Visualize 14 points with their angles (if provided).

preprocess_floor_plan must be called beforehand, as this funciton reads from the npz files it writes. Saves normalized data (ready f