Readspeaker menu

Top 5% Ranking in Google Landmark Retrieval and Recognition Challenges

We are proud to announce that our Ph.D. student Björn Barz achieved outstanding results in two recent competitions in the field of content-based image retrieval. Both competitions were organized by Google and conducted via the data science platform Kaggle.

Given an image of one of over 15,000 landmarks from all over the world, finding similar images of the same landmark in a database of more than one million images was the goal of the Google Landmark Retrieval Challenge, where Björn reached the 10th place among a total of 218 participants (top 5%).

The Google Landmark Recognition Challenge, on the other hand, asked the competing methods to classify the given image, i.e., provide the name of the landmark instead of similar images. In this competition, Björn reached the 12th place among a total of 483 participants (top 3%).

The method used for the Retrieval Challenge will be presented at the Large-Scale Landmark Recognition Workshop at the IEEE Conference on Computer Vision and Pattern Recognition, which will take place in Salt Lake City in the end of July.

previous | next