must_downloads.txt

1. Download the 'alexnet.caffemodel' from 'http://dl.caffe.berkeleyvision.org/bvlc_alexnet.caffemodel', and copy to 'DIR/VPR-Bench/VPR_Techniques/AlexNet_VPR/alexnet/'.

2. Download and extract the 'calc.caffemodel' from 'http://udel.edu/~nmerrill/calc.tar.gz', and copy to 'DIR/VPR-Bench/VPR_Techniques/CALC/model/'. 

3. Download the 'AmosNet.caffemodel' from 'https://goo.gl/6xtjwD', and copy to 'DIR/VPR-Bench/VPR_Techniques/AmosNet/'.

4. Download the 'HybridNet.caffemodel' from 'https://goo.gl/kF6nQh', and copy to 'DIR/VPR-Bench/VPR_Techniques/HybridNet/'.

5. Download the tensorflow checkpoints for NetVLAD from 'http://rpg.ifi.uzh.ch/datasets/netvlad/vd16_pitts30k_conv5_3_vlad_preL2_intra_white.zip', extract and copy its content to the 'DIR/VPR-Bench/VPR_Techniques/NetVLAD/checkpoints/'.

6. Download the 'alexnet_places365.caffemodel' from 'http://places2.csail.mit.edu/models_places365/alexnet_places365.caffemodel' and copy to '/DIR/VPR-Bench/VPR_Techniques/RegionVLAD/AlexnetPlaces365/'.

7. Download the 'variation_quantified_datasets' folder from 'https://surfdrive.surf.nl/files/index.php/s/65AMTSuxxo2siJT' and extract it in the VPR-Bench folder. 

8. (optional) There are 12 datasets used in our VPR-Bench paper, you can download these datasets (or necessary information to get them) from this link 'https://surfdrive.surf.nl/files/index.php/s/sbZRXzYe3l0v67W'. Since the corridor dataset is already available in this GitHub repo, you do not need to have downloaded these 12 datasets to get the VPR-Bench to work but obviously after the first trial, you would want to see performance on other datasets as well and this is where you need to go then. 

9. (optional) Download precomputed matching information of the 10 techniques on the 12 datasets used in our work from 'https://surfdrive.surf.nl/files/index.php/s/ThIgFycwwhRCVZv', extract and copy it to the 'VPR-Bench/precomputed_matches/' folder.