socc-plotter 0.2.0

Awesome socc_plotter created by AdityaNG

socc_plotter

Semantic Occupancy 3D Plotter. This is the plotter made by the SOccDPT project to create fancy 3D visuals. You can use this for your own AV or Robotics visualization!

Install it from PyPI

pip install socc_plotter

Usage

The socc_plotter works on a callback mechanism since the GUI must be run on the main thread.

from socc_plotter.plotter import Plotter
import time

def callback(plot: Plotter):
    time.sleep(1)
    print("in callback")
    graph_region = plot.graph_region

    points = np.array([[1, 0, 0]])
    colors = np.array([[1, 1, 1]])

    graph_region.setData(pos=points, color=colors)

plotter = Plotter(
    callback=callback,
)
plotter.start()

NuScenes Demo

Start by downloading the NuScenes mini datset

mkdir -p data/nuscenes/
cd data/nuscenes/
wget -c https://www.nuscenes.org/data/v1.0-mini.tgz
tar -xf v1.0-mini.tgz

Install a few dependencies for the demo

pip install nuscenes-devkit==1.1.10
pip install transformers torch torchvision timm accelerate general_navigation

Run the demo

$ python -m socc_plotter
#or
$ socc_plotter

Development

Read the CONTRIBUTING.md file.

Cite

Cite our work if you find it useful

@article{NG2024SOccDPT,
  title={SOccDPT: 3D Semantic Occupancy from Dense Prediction Transformers trained under memory constraints},
  author={NG, Aditya},
  journal={Advances in Artificial Intelligence and Machine Learning},
  volume={ISSN: 2582-9793, Source Id: 21101164612},
  year={2024},
  url={https://www.oajaiml.com/}
}

TODO

• Demo
  • RGB Frame
  • Depth perception
  • Semantic segmentation
  • NuScenes Calibration
  • NuScenes Vehicle trajectory
  • Semantic Occupancy Grid
• Ensure demo dependencies are seperate from the module
• Demo is to prompt the user to install dependencies
• Demo is to auto download NuScenes and unarchive it
• Test Cases
• PiPy deployment