Supplementary Material for Unsupervised Adaptation from Repeated Traversals for Autonomous Driving S1 Implementation Details

The parameters that we used in this work were β = 0.333, and N We include an ablation table for different values of β in Table S1. For the focal loss, we set α = 0.25 and γ = 2.0 which are the default values. We selected the best hyperparameters based on the performance on KITTI Lyft and used the same hyperparameters for the rest of the settings. We show results experiementing with different β parameters. We include additional evaluations on the Lyft dataset.