QVPR / Patch-NetVLAD

Code for the CVPR2021 paper "Patch-NetVLAD: Multi-Scale Fusion of Locally-Global Descriptors for Place Recognition"
MIT License
531 stars 76 forks source link

Add image crop transform #78

Open f-fl0 opened 1 year ago

f-fl0 commented 1 year ago

First of all thank you for sharing the implementation for Patch-NetVLAD!

This PR introduces the imagecrop entry in the configuration to allow the user to specify a custom region of interest in the input image. To enable it, one can just add the following entry in the feature_extract and/or feature_match section(s) of the configuration file:

imagecrop = top,left,bottom,right