Code for Adversarial Feature Disentanglement for Place Recognition Across Changing Appearance.

Dependencies

We develop our code under anaconda, with following depenencies (not listed fully):

• pytorch 1.1.0
• torchvision 0.3.0
• cuda 9.0
• tensorboardx 1.7
• toolz 0.9
• matplotlib 3.0.3
• hdf5storage 0.1.15
• numpy 1.16.4
• scikit-learn 0.21.1
• pyyaml 5.1
• py-opencv 3.4.2
• attrdict 2.0.1
• latex? (remove it later)

Run

MNIST toy case:

python main.py config/mnist.yaml

Place recognition:

python main.py config/pr.yaml