INTRODUCTION
One task, many names
Find the corresponding point across time, viewpoints, sensors, or object instances. Each depends on the pairs used for training.
Follow the evidence
Each topic opens its own page.Slide 2
One task, many names
Open evidence →
Slide 3Under the hood, matching means comparing features
Open evidence →
Slide 4Supervised matching needs pairs that barely exist
Open evidence →
Slide 5But the motion is not noise — it has structure
Open evidence →
Slide 6The motion that a generator produces predicts transfer better than the realism of its renderings.
Open evidence →
Slide 7From a data constraint to a design method
Open evidence →