general-navigation 0.6.0

Awesome general_navigation created by AdityaNG

General Navigation

General Navigation Models based on GNM, ViNT, NoMaD as a pytorch repo for quick and easy deployment.

Install it from PyPI

PyPi Link

Install our project from pip and quickly get started by trying it out on your own test video!

pip install general_navigation
python3 -m general_navigation --media media/test.mp4

If you want to connect with the Carla simulator, you will also need to seperately install carla

pip install carla==0.9.15  # Linux and Windows
pip install carla==0.9.5  # Mac

Usage

Documentation Link

Creating a pytorch instance of the model

from general_navigation.models.factory import (
  get_default_config, get_model, get_weights
)

config = get_default_config()
model = get_model(config)
model = get_weights(config, model, device)

Using the command line tool for inference on a video

$ python3 -m general_navigation --help
usage: general_navigation [-h] [--device {auto,cuda,cpu}] [--media MEDIA]

options:
  -h, --help            show this help message and exit
  --device {auto,cuda,cpu}, -d {auto,cuda,cpu}
  --media MEDIA, -m MEDIA
                        File path, use camera index if you want to use the webcam
$ python3 -m general_navigation --media media/test.mp4

Using the command line tool for working with Carla simulator

$ python3 -m general_navigation.carla_run

TODO

• Import models from visualnav-transformer
• Script to use specified video camera to feed video into the models
• Visualize model's trajectory output
• Arguments to CLI
• Auto download model weights from google drive
• Demo video
• PyPi release
• Example usage
• Carla Integration
• MPC implementation
• Visualize MPC reconstructed trajectory
• Visualize all of the model's trajectories
• Visualize Steering angle
• Fix scaling issue
• Intrinsic matrix as argument
• Documentation: mkdocs gh-deploy
• Linting fixes
• MyPy testing
• Add test cases for code coverage

Development

Read the CONTRIBUTING.md file.

References

• visualnav-transformer
• diffusion_policy