Skip to content

mie-lab/traffic4cast

Repository files navigation

Traffic4cast–Traffic Map Movie Forecasting

by Team MIE-Lab: Henry Martin, Ye Hong, Dominik Bucher, Christian Rupprecht, René Buffat

Documentation

Our documentation is now available!

Howto

Train a model

  • Copy the competition raw data into data_raw folder such that the top-level city folders are visible.
  • Training configuration can be changed in the config_unetfile (e.g., choose a city).
  • Run the training_unet.py file in the traffic4castfolder
  • The training is logged via tensorboard in the runs folder

Create a submission with the pretrained models

  • Download the pretrained models using this link
  • Store the 3 top-level model folders (e.g., UNet_Berlin) in the runs folder (you can overwrite the existing ones)
  • Run the example_create_3modelsubmission.py script.

This work was extended in a master thesis. code

About

Submission to the iarai traffic4cast competition

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages