DAN-SuperPoint: Self-Supervised Feature Point Detection Algorithm with Dual Attention Network

Abstract

In view of the poor performance of traditional feature point detection methods in low-texture situations, we design a new self-supervised feature extraction network that can be applied to the visual odometer (VO) front-end feature extraction module based on the deep learning method. First, the network uses the feature pyramid structure to perform multi-scale feature fusion to obtain a feature map containing multi-scale information. Then, the feature map is passed through the position attention module and the channel attention module to obtain the feature dependency relationship of the spatial dimension and the channel dimension, respectively, and the weighted spatial feature map and the channel feature map are added element by element to enhance the feature representation. Finally, the weighted feature maps are trained for detectors and descriptors respectively. In addition, in order to improve the prediction accuracy of feature point locations and speed up the network convergence, we add a confidence loss term and a tolerance loss term to the loss functions of the detector and descriptor, respectively. The experiments show that our network achieves satisfactory performance under the Hpatches dataset and KITTI dataset, indicating the reliability of the network.

An Efficient Indoor Localization Based on Deep Attention Learning Model

“…ORB [10], etc.) or learned features (such as SuperPoint [11], ASLFeat [20], or SeqNet [21]) can be used as descriptors. Clustering algorithms such as [22,23] can play an essential role in improving the accuracy of feature-based techniques.…”

“…To find matches between images, the global descriptors are searched. A descriptor of an image can be either a handcrafted feature (e.g., Scale Invariant Feature Transform (SIFT) [7,8], Speeded Up Robust Features (SURF) [9], or Oriented FAST and Rotated BRIEF (ORB) [10]) or a learned feature (e.g., SuperPoint [11]). Although feature-based methods are powerful in many situations, it still faces challenges when there are less texture, repetitive structures, and insufficient matching features [12,13].…”

RU-SLAM: A Robust Deep-Learning Visual Simultaneous Localization and Mapping (SLAM) System for Weakly Textured Underwater Environments

“…On the other hand, scholars have tried to incorporate deep features into SLAM [ 10 , 11 , 12 ]. Methods based on deep learning possess the capability to process full-size images and simultaneously compute pixel-level keypoints and their descriptors.…”

“…To enhance the system stability, ref. [ 12 ] introduced a novel deep learning SLAM system that utilizes self-supervised learning to optimize the network. While deep learning-based SLAM methods often outperform traditional approaches in complex environments, they frequently require substantial labeled datasets for training.…”