Skip to content
Learning Matchable Image Transformations
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Matchable Image Transformations for Long-term Visual Localization

System overview

Code to accompany our paper "Learning Maximally Matchable Image Transformations for Long-term Metric Visual Localization".


  • numpy
  • matplotlib
  • pytorch 1.1 + torchvision
  • Pillow
  • progress (for progress bars in train/val/test loops)
  • pyviso2 (included)
  • tensorboardX


  • - Top-level script for training and testing models using the Virtual KITTI dataset.
  • - Top-level script for training and testing models using the UTIAS In The Dark dataset.
  • matchability/ - Main codebase.
    • datasets/ - Modules for loading and pre-processing the Virtual KITTI and UTIAS In The Dark datasets for use with PyTorch.
    • - General training, validation and testing loops.
    • - Definitions for matcher proxy and image transformation models.
    • - General-purpose neural network classes, specialized in
    • - Defines a container class to store configuration parameters and provides common defaults.
    • - Custom image/tensor transformations.
    • - Miscellaneous utility functions.
  • scripts/ - Scripts for producing figures and videos used in the paper.
You can’t perform that action at this time.