Skip to content

Latest commit

 

History

History
59 lines (35 loc) · 1.23 KB

File metadata and controls

59 lines (35 loc) · 1.23 KB

1. Shape Processing

1.1. Download ModelNet40

The ModelNet40 dataset can be downloaded from here.

Then create a soft link by running:

ln -s /path/to/download/modelnet40_normal_resampled /path/to/this/directory/

1.2. Process the shapes

Run command:

python modelnet40_tools.py

2. Virtual Scene Generation (Point-version)

2.1. Download Matterport3D

The Matterport3D dataset can be downloaded from here.

2.2. Generate augmentation information

Run commands below:

cd /path/to/BackToReality/detection/Votenet/matterport

python matterport_detection_dataset.py

mv scans_toadd_scarse.npy /path/to/BackToReality/data_generation/Matterport3D/CONFIG

2.3. Generate the positions for shapes

Run commands below:

cd /path/to/BackToReality/data_generation/Matterport3D

python matterport_scene_synthesis.py

mv augment_random_positions_matterport /path/to/BackToReality/detection/Votenet/matterport

2.4. Use the positions to generate virtual scenes

Run commands below:

cd /path/to/BackToReality/detection/Votenet/matterport

python batch_load_matterport_data_virtual.py