Code for learning local image descriptors with autoencoedrs and variational autoencoders

software.png
Description:

Repository for reproducing results of our paper “Efficient Local Image Descriptors Learned with Autoencoders” that is currently under review in IEEE Access. It contains code for training autoencoders and variational autoencoders, performing a hyperparameter sweep over different parameters as described in the paper, and for automatically evaluating the descriptor’s performance, both based on the established HPatches benchmark and based on our proposed approximate evaluation method.

 

More Information:
Publications:
  • N. Zizakic and A. Pizurica, "Efficient Local Image Descriptors Learned with Autoencoders", IEEE Access, 2021, (Submitted)