Garage Points

Today I learned that parking spaces so important to self driving car research that there is now a dataset just for garages called "LetsGo". The paper for it can be found here.


There’s a few interesting things about garages when you think about it for self driving use-cases.

  • For a bunch of cars, the parking garage will be both the starting point as well as the destination of a journey.
  • Lighting conditions can are usually rather poor.
  • Paths contain spirals, which are very rare outside of garages.
  • The walls typically have very little texture to go on because it’s all concrete most of the time.

To aid in all this, the researchers collected a LiDAR dataset as well as fisheye imagery.


It hadn’t occured to me before, but it totally makes sense to have an effort surrounding garages for self driving. Especially in the US, where these garages can get insanely big, you really want to make sure that your dataset has garages covered.