1) Install PyViz3D (more lightweight than other tools such as Mayavi or Open3D).
pip3 install pyviz3d
2) Export the predictions of the trained-model. For example:
python3 tools/test.py configs/scannetv2/isbnet_scannetv2.yaml head_isbnet_scannetv2 --out results/isbnet_scannetv2_val
3) The results folder is structured as follows.
ISBNet
├── results
│ ├── isbnet_scannetv2_val
│ │ ├── pred_instance
│ │ │ ├── predicted_masks
│ │ │ │ ├── scene0011_00_001.txt
│ │ │ │ ├── scene0011_00_002.txt
│ │ │ │ ├── ...
│ │ │ │ ├── scene0011_00_100.txt
│ │ │ │ ├── scene0011_01_001.txt
│ │ │ │ ├── scene0011_01_002.txt
│ │ │ │ ├── ...
│ │ │ ├── scene0011_00.txt
│ │ │ ├── scene0011_01.txt
│ │ │ ├── ...
│ │ │ ├── scene0704_01.txt
4) Visualize the result:
python3 visualization/vis_scannetv2.py --data_root dataset/scannetv2 --scene_name scene0011_00 --prediction_path results/isbnet_scannetv2_val --task inst_pred
5) Follow the instructions on the terminal:
# open a new terminal and type:
cd ISBNet/visualization/pyviz3d; python -m http.server 6008
# open on your browser to see the result:
http://0.0.0.0:6008
6) You can also follow the instructions from SoftGroup to visualize the results using Open3D (visualization.py).