Skip to content

Latest commit

 

History

History
 
 

rerun_visualization

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
docs
docs
 
 
README.md
README.md
 
 
visualize_3d_model.py
visualize_3d_model.py
 
 

README.md

tools/rerun_visualization

Visualize with T4dataset.

  • Support priority: Tier S
  • Supported feature
    • T4dataset 3D ground truth visualization
    • T4dataset 3D inference visualization
    • T4dataset 2D ground truth visualization
    • T4dataset 2D inference visualization
    • T4dataset 3D ground truth analysis
    • T4dataset 3D inference analysis
  • Other supported feature
    • Add unit test

Install and setup

  • Install on native environment
pip install rerun-sdk==0.17.0

1. 3D Visualization

Visualize 3D detection with T4dataset.

  • Run docker
docker run -it --rm --gpus 'all,"capabilities=compute,utility,graphics"' --shm-size=64g --name awml -v $PWD/:/workspace -v $HOME/local/dataset:/workspace/data --net host -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY autoware-ml
  • Stand another terminal and Run on your native environment
rerun

Visualize of the results from 3D detection trained model

  • Run visualization scripts in docker environment
  • Visualize ground truth
python tools/rerun_visualization/visualize_3d_model.py \
{config_file} \
{product_config} \
--checkpoint {checkpoint_file} \
--fix-rotation --split test --bbox-score 0.4 --objects ground_truth --image-num 6
  • Visualize prediction result
python tools/rerun_visualization/visualize_3d_model.py \
{config_file} \
{product_config} \
--checkpoint {checkpoint_file} \
--fix-rotation --split test --bbox-score 0.4 --objects prediction --image-num 6
  • (TBD) Visualize prediction result with ground truth
python tools/rerun_visualization/visualize_3d_model.py \
{config_file}\
{product_config} \
--checkpoint {checkpoint_file} \
--fix-rotation --split test --bbox-score 0.4 --objects prediction_with_gt --image-num 6
  • For example, visualize XX1 with TransFusion-L
python tools/rerun_visualization/visualize_3d_model.py \
work_dirs/pretrain/transfusion/transfusion_lidar_pillar_second_secfpn_1xb1_90m-768grid-t4xx1.py \
autoware_ml/configs/detection3d/dataset/t4dataset/xx1.py \
--checkpoint work_dirs/pretrain/transfusion/epoch_50.pth \
--fix-rotation --split test --bbox-score 0.1 --objects prediction --image-num 6

(TBD) Visualize of the results from info file

  • Run visualization scripts in docker environment
python tools/rerun_visualization/visualize_3d_pkl.py {pkl_file}
--fix-rotation --split test --bbox-score 0.4 --objects prediction --image-num 6

2. Analysis for 3D detection

  • Analysis for dataset
TBD
  • Analysis for inference results
TBD

3. 2D visualization

TBD

Reference