Tanks and Temples

Tank and Temples is a benchmark that uses Lidar point clouds as ground truth for benchmarking the quality of image-based 3-d reconstruction algorithms. The point clouds are stored as .ply files. Here we show how to load and visualize these point clouds.

Download Truck.ply (392 MB) [link].

Import required Python packages. Here we use the plyfile Python package to read .ply files.

>>> import pptk
>>> import numpy as np
>>> import plyfile

Note

If pip install plyfile does not work, simply save a local copy of plyfile.py from plyfile’s github page.

Read vertices in Truck.ply.

>>> data = plyfile.PlyData.read('Truck.ply')['vertex']

Use per-vertex attributes to make numpy arrays xyz, rgb, and n.

>>> xyz = np.c_[data['x'], data['y'], data['z']]
>>> rgb = np.c_[data['red'], data['green'], data['blue']]
>>> n = np.c_[data['nx'], data['ny'], data['nz']]

Visualize.

>>> v = pptk.viewer(xyz)
>>> v.attributes(rgb / 255., 0.5 * (1 + n))

Use [ and ] to toggle between attributes.

truck_rgb truck_n
The Truck.py point cloud from Tanks and Temples visualized using pptk.viewer(). Points are colored by RGB (left) and normal (right)

The above procedure can be repeated for the other point clouds in the dataset.