Added pitts30k as an option for PCA training

QVPR / Patch-NetVLAD

Code for the CVPR2021 paper "Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition"

MIT License

525 stars 74 forks source link

Added pitts30k as an option for PCA training #39

Closed StephenHausler closed 2 years ago

StephenHausler commented 2 years ago

Added new --dataset_choice arg, which allows a user to train PCA using either mapillary (default) or pitts30k. Code has been successfully tested. In response to Issue #37.

StephenHausler commented 2 years ago

What new file? I've coded it differently to how we discussed, I'm re-using the existing patchnetvlad/tools/datasets.py to load the pitts pca data, instead of creating a replica of MSLS.

StephenHausler commented 2 years ago

Requested changes complete