Presentation slides (55MB pdf)
Local feature matching is one of the cornerstones of “classical” computer vision. Despite the recent advances of deep learning, a crucial question still remains, which is whether the learned local feature methods can outperform the classical methods. Recent results have shown that classical non-deep learning methods are still very competitive and can even outperform state-of-the-art deep learning methods in specific cases.
This tutorial will aim to present a comprehensive overview of the classical local feature methods, explore the current solutions that are based on deep learning, and finally examine the drawbacks and strengths of both sides in order to provide researchers with pointers about current challenges and inspire future work.