FPC-Net: Revisiting SuperPoint with Descriptor-Free Keypoint Detection via Feature Pyramids and Consistency-Based Implicit Matching
This work addresses memory efficiency in localization systems for computer vision applications, but it is incremental as it builds on existing keypoint detection methods like SuperPoint.
The paper tackles the problem of interest point matching in computer vision by introducing a descriptor-free method that inherently associates points during detection, eliminating descriptors and drastically reducing memory usage, though with marginally lower matching accuracy than conventional approaches.
The extraction and matching of interest points are fundamental to many geometric computer vision tasks. Traditionally, matching is performed by assigning descriptors to interest points and identifying correspondences based on descriptor similarity. This work introduces a technique where interest points are inherently associated during detection, eliminating the need for computing, storing, transmitting, or matching descriptors. Although the matching accuracy is marginally lower than that of conventional approaches, our method completely eliminates the need for descriptors, leading to a drastic reduction in memory usage for localization systems. We assess its effectiveness by comparing it against both classical handcrafted methods and modern learned approaches.