Reproducing the result from Oxford RobotCar dataset

[yunjinli]

Hi Alex,
I hope you are doing well! I just have some questions regarding to reproducing the result from the TABLE III TOP-1% RECALLS FROM OXFORD ROBOTCAR DATASET in your paper. Can I just use the pretrained manydepth weight that you provided and also the pretrained weight of your dual encoder to reproduce the result or I somehow need to retrain the dual encoder with Oxford RobotCar dataset? I look forward to your reply, thanks!

[alexjunholee]

Hi @yunjinli,

You'll need the fine-tuned weights for every module (including the depth estimation module) to evaluate them in the Oxford dataset. Unfortunately, we're not currently providing it at this moment. Thanks for paying attention to our work.