npy2pointcloud

Convert Rohbau3D .npy point cloud files to standard formats (PLY, PCD, LAS).

Supported Formats

Format	Extension	Library	Colors	Normals	Intensity	Notes
PLY	.ply	Open3D	Yes	Yes	--	Binary format, widely supported
PCD	.pcd	Open3D	Yes	Yes	--	Point Cloud Library native format
LAS	.las	laspy	Yes (16-bit)	Yes (extra dims)	Yes	LAS 1.4, point format 7

Rohbau3D Input Format

Each scene directory contains:

File	Shape	Description
coord.npy	(N, 3) float	XYZ coordinates (required)
color.npy	(N, 3) uint8	RGB values
intensity.npy	(N, 1) float	Laser reflectance
normal.npy	(N, 3) float	Surface normals

Installation

pip install -e .

Usage

Convert a single scene

npy2pointcloud convert -i /path/to/scene -o output.ply -f ply
npy2pointcloud convert -i /path/to/scene -o output.las -f las
npy2pointcloud convert -i /path/to/scene -o output.pcd -f pcd

Show point cloud info

npy2pointcloud info -i /path/to/scene

Example output:

Source:     /data/rohbau3d/scene_001
Points:    10,452,301
XYZ range: x=[-12.345, 45.678]  y=[-8.901, 23.456]  z=[0.123, 15.789]
Colors:    yes
Intensity: yes
Normals:   yes

Batch convert an entire dataset

# Mirror directory structure
npy2pointcloud batch -i /path/to/dataset -o /path/to/output -f ply

# Flatten into a single directory
npy2pointcloud batch -i /path/to/dataset -o /path/to/output -f las --flatten

Python API

from npy2pointcloud.loader import load_scene
from npy2pointcloud.converter import convert

data = load_scene("/path/to/scene")
print(data.summary())

convert(data, "output.ply", "ply")
convert(data, "output.las", "las")
convert(data, "output.pcd", "pcd")

Development

pip install -e .
pip install pytest
pytest tests/ -v