Large-Scale Landmark Recognition: A Challenge

LandmarksCVPRW18

This workshop fosters research on image retrieval and landmark recognition by introducing a novel large-scale dataset, together with evaluation protocols. More details of the challange and the dataset can be found here

Recognition Challenge

https://www.kaggle.com/c/landmark-recognition-challenge

Label famous (and not-so-famous) landmarks in image.

Retrieval Challenge

https://www.kaggle.com/c/landmark-retrieval-challenge

Given an image, can you find all of the same landmarks in a dataset?


Workshop Schedule

Challenge Begins

Challenge is held on Kaggle

Feb. 2 2018

Challenge Deadline

Results are released

May 27 2018

Workshop

Held with CVPR’18

Jun. 18 2018, 13:30 - 17:30, MDT

Workshop (Location: Room 251 - BC)

Welcome Remark

Andre Araujo
13:30 - 13:50, MDT

Invited Talk 1

Josef Sivic (INRIA / CTU) (Hosted by Ondra Chum)
Learnable Representations for Estimating Visual Correspondence

13:50 - 14:30, MDT

Challenge Overview

Andre Araujo and Tobias Weyand
14:30 - 14:50, MDT

Recognition Challenge Winner Presentations

Sirius, VRG Prage and Kohei (Hosted by Torsten Sattler)
14:50 - 15:30, MDT

Coffee Break

15:30 - 16:00, MDT

Retrieval Challenge Winner Presentations

CVSSP & Visual Atoms, Layer 6 AI, and SevenSpace (Hosted by Jack Sim)
16:00 - 16:40, MDT

Invited Talk 2

Herve Jegou (FAIR) (Hosted by Ondra Chum)
Trends in Large-Scale Similarity Search

16:40 - 17:20, MDT

Close Remark

Tobias Weyand
17:20 - 17:30, MDT

Challenge Participant Slides

Link to Slides

Invited Speakers

Josef Sivic

Senior researcher at Inria and a principal investigator at Czech Technical University in Prague

Learnable Representations for Estimating Visual Correspondence

Finding visual correspondence is one of the fundamental image understanding problems. It is a challenging task due to strong appearance differences between the corresponding scene elements caused by, for example, changes in camera viewpoint and illumination, intra-class variation, or scenes changes over time. In this talk, I will discuss our recently developed learnable representations for visual correspondence tasks with applications to category-level object alignment, observer localization in large-scale indoor environments with textureless areas and repetitive patterns, and place recognition across day/night illumination or changes of weather and seasons.

Herve Jegou

Research Lead with Facebook AI Research

Trends in Large-Scale Similarity Search

In this talk, I will present some recent trends in the area of similarity search. I will first review the techniques that are routinely employed to index and search billions of image descriptors, such as local features or more global representations extracted from CNN architectures. Then I will present recent advances based on graph-descent algorithms, as well as emerging methods for indexing with neural networks.


Organizer

Bohyung Han

Associate Professor, Seoul National University (Primary Contact)

Andre Araujo

Software Engineer, Google (Primary Contact)

Shih-Fu Chang

Professor, Columbia University

Ondrej Chum

Associate Professor, Czech Technical University

Torsten Sattler

Senior Researcher, ETH

Jack Sim

Software Engineer, Google

Giorgos Tolias

Postdoc, Czech Technical University

Tobias Weyand

Software Engineer, Google

Xu Zhang

Postdoc, Columbia University

© 2018 Landmarks18

We thank Jalpc for the jekyll template