Need help with revisitop?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

filipradenovic
136 Stars 36 Forks 37 Commits 6 Opened issues

Description

Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking

Services available

!
?

Need anything else?

Contributors list

No Data

Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking

We revisit and address issues with Oxford 5k and Paris 6k image retrieval benchmarks. New annotation for both datasets is created with an extra attention to the reliability of the ground truth and three new protocols of varying difficulty are introduced. We additionally introduce 15 new challenging queries per dataset and a new set of 1M hard distractors.

This package provides support in downloading and using the new benchmark.

MATLAB

Tested with MATLAB R2017a on Debian 8.1.

Process images

This example script first downloads dataset images and the revisited annotation files. Then, it describes how to: read and process database images; read, crop and process query images: ```

exampleprocessimages

Similarly, this example script first downloads one million images from the revisited distractor dataset (this can take a while). Then, it describes how to read and process images.
exampleprocessdistractors ```

Evaluate results

Example script that describes how to evaluate according to the revisited annotation and the three protocol setups: ```

exampleevaluate

It automatically downloads dataset images, the revisited annotation file, and example features (R-[37]-GeM from the paper) to be used in the evaluation.
The final output should look like this (depending on the selected
testdataset

):
roxford5k: mAP E: 84.81, M: 64.67, H: 38.47 roxford5k: [email protected][1 5 10] E: [97.06 92.06 86.49], M: [97.14 90.67 84.67], H: [81.43 63.00 53.00]
or
rparis6k: mAP E: 92.12, M: 77.20, H: 56.32 rparis6k: [email protected][1 5 10] E: [100.00 97.14 96.14], M: [100.00 98.86 98.14], H: [94.29 90.29 89.14] ```

Python

Tested with Python 3.5.3 on Debian 8.1.

Process images

This example script first downloads dataset images and the revisited annotation files. Then, it describes how to: read and process database images; read, crop and process query images: ```

python3 exampleprocessimages

Similarly, this example script first downloads one million images from the revisited distractor dataset (this can take a while). Then, it describes how to read and process images.
python3 exampleprocessdistractors ```

Evaluate results

Example script that describes how to evaluate according to the revisited annotation and the three protocol setups: ```

python3 exampleevaluate

It automatically downloads dataset images, revisited annotation file, and example features (R-[37]-GeM from the paper) to be used in the evaluation.
The final output should look like this (depending on the selected
testdataset

):
roxford5k: mAP E: 84.81, M: 64.67, H: 38.47 roxford5k: [email protected][ 1 5 10] E: [97.06 92.06 86.49], M: [97.14 90.67 84.67], H: [81.43 63. 53. ]
or
rparis6k: mAP E: 92.12, M: 77.2, H: 56.32 rparis6k: [email protected][ 1 5 10] E: [100. 97.14 96.14], M: [100. 98.86 98.14], H: [94.29 90.29 89.14] ```

Related publication

@inproceedings{RITAC18,
 author = {Radenovi\'{c}, F. and Iscen, A. and Tolias, G. and Avrithis, Y. and Chum, O.},
 title = {Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking},
 booktitle = {CVPR},
 year = {2018}
}

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.