KLN: A Deep Neural Network Architecture for Keypoint Localization

Localization of keypoints on pixel-level precision is an essential step for stitching panoramic images because the keypoints are matching, and their locations are used for computing stitching transformation. We recall the main standard computer vision techniques for keypoint localization and focus on the precise localization. We design a neural network architecture containing an encoder, a latent representation handler, and a decoder, where the encoder is motivated by SIFT. In contrast to domain-agnostic neural network architectures, the developed encoder reflects the scale-space construction as well as the difference of Gaussians estimation used in SIFT. In the benchmark, we show that our architecture has a higher number of keypoints localized with pixel-level precision than other standard and neural network-based approaches.