Evaluation on the NCLT dataset

[BIT-MJY]

Hello @Suoivy,

Thanks for your nice LPD-Net which still works well today!

Have you evaluated LPD-Net on the NCLT dataset? While I train LPD-Net with one sequence of the NCLT dataset and test it on the other seqs, LPD-Net does not perform well enough as on Robotcar dataset. Thus I doubt that some details of my training process went wrong. Could please provide some reference in terms of the performance of LPD-Net on the NCLT dataset, such as recall@1 or AUC on some sequences?

Thanks very much!