Releasing ground truth for Phototourism and PragueParks test set
We have released groud truth test set used for the 2021 challenge, and the evaluation server has been closed. You are welcome to download the ...
News
Announcing the results of the 2021 Image Matching Challenge!
We are happy to announce the winners of this year's Image Matching Challenge: Restricted keypoints category Dongli Tan, Xingyu Chen, Ruixin Zhang, Kai Zhao, Xuehui ...
SimLocMatch has been announced!
Synthetic dataset image matching challenge is online now! It will be held at the Image Matching Workshop @ CVPR 2021. Please visit https://simlocmatch.com/ to submit ...
A new tutorial by Dmytro on custom matcher
Dmytro, one of our co-organizers, wrote a new tutorial on how to create a submission with custom match files. It can be found here: https://ducha-aiki.github.io/wide-baseline-stereo-blog/2021/05/27/submitting-to-IMC2021-with-custom-matcher.html ...
Minor modification to the rules on descriptor size
In order to allow submissions of papers/methods that do not have a clear feature size, such as those relying on dense feature maps, we have ...
A new submission example
Dmytro, one of our co-organizers, took the time to write a thorough tutorial on how to use the benchmark to tune your method and create ...
Introducing the 2021 Challenge
We are pleased to announce the third edition of the Image Matching Challenge, which introduces two new datasets . In addition to the "Phototourism" dataset, ...
Announcing the results for the 2020 Challenge
We are pleased to announce that the 2020 edition of the Image Matching Challenge is over, and it was a success. We received over 100 ...
Bugfixes and leaderboard updates
We would like to announce some updates: In order to prevent possible collisions due to multiple submissions possibly using the same method name or label ...
Postponing challenge deadline
With CVPR likely moving towards a virtual format, there is no need to have a deadline so far from the conference. We have thus moved ...
Removing United States Capitol
It has been brought to our attention that some of the images from United States Capitol are not be correctly registered, and in fact show ...
Running the benchmark in a scalable way
The benchmark relies on the Slurm job scheduler , which allows users to queue jobs (with dependencies) on multi-user clusters. We made this decision for ...
Introducing the 2020 Image Matching Challenge
It is our pleasure to announce the second edition of the Image Matching Challenge, held in conjunction with the CVPR 2020 workshop on Image Matching: ...