vis.md

Visualization

First, run the following command:

python tools/visualize.py PATH_TO_CONFIG --checkpoint PATH_TO_CHECKPOINT

This will generate a folder structured like this:

viz/
├── camera-0
│   ├── 1531281439800013-*.png
│   └── ...
├── camera-1
│   ├── 1531281439800013-*.png
│   └── ...
├── camera-2
│   └── ...
├── camera-3
│   └── ...
├── camera-4
│   └── ...
├── camera-5
│   └── ...
├── lidar
│   └── ...
└── occ
    ├── 1531281439800013-*.npy
    ├── 1531281439800013-*_gt.npy
    └── ...

• Files with suffix _gt.npy represent the ground truth.
• Files with suffix .npy represent the model predictions.

Prerequisites

Before visualizing the occupancy results, please ensure the following dependencies are installed:

pip install imageio
pip install vtk==9.3.1
pip install mayavi==4.8.2 --no-cache-dir --verbose --no-build-isolation
pip install PyQt5

Occupancy Visualization

To visualize the predicted occupancy, use the following command:

python tools/vis_occ3d.py viz/occ/1531281439800013-*.npy

• Replace the file name with the actual .npy file name. Do not use wildcards (*) in the command.