In this paper, we propose DeepSLAM, a novel unsupervised deep learning-based visual Simultaneous Localization and Mapping (SLAM) system. The DeepSLAM training is fully unsupervised since it only requires stereo imagery instead of annotating ground-truth poses. Its testing takes a monocular image sequence as the input. Therefore, it is a monocular SLAM paradigm. DeepSLAM consists of several essential components, including Mapping-Net, Tracking-Net, Loop-Net and a graph optimization unit. Specifically, the Mapping-Net is an encoder and decoder architecture for describing the 3D structure of the environment while the Tracking-Net is a Recurrent Convolutional Neural Network (RCNN) architecture for capturing the camera motion. The Loop-Net is a pre-trained binary classifier for detecting loop closures. DeepSLAM can simultaneously generate pose estimate, depth map and outlier rejection mask. We evaluate its performance on various datasets, and find that DeepSLAM achieves good performance in terms of pose estimation accuracy, and is robust in some challenging scenes.

The exceptional performance of general-purpose large models has driven various industries to focus on developing domain-specific models. However, large models are not only time-consuming and labor-intensive during the training phase but also have very high hardware requirements during the inference phase, such as large memory and high computational power. These requirements pose considerable challenges for the practical deployment of large models. As these challenges intensify, model compression has become a vital research focus to address these limitations. This paper presents a comprehensive review of the evolution of model compression techniques, from their inception to future directions. To meet the urgent demand for efficient deployment, we delve into several compression methods—such as quantization, pruning, low-rank decomposition, and knowledge distillation—emphasizing their fundamental principles, recent advancements, and innovative strategies. By offering insights into the latest developments and their implications for practical applications, this review serves as a valuable technical resource for researchers and practitioners, providing a range of strategies for model deployment and laying the groundwork for future advancements in model compression.

This paper presents an indoor relocalization system using a dual-stream convolutional neural network (CNN) with both color images and depth images as the network inputs. Aiming at the pose regression problem, a deep neural network architecture for RGB-D images is introduced, a training method by stages for the dual-stream CNN is presented, different depth image encoding methods are discussed, and a novel encoding method is proposed. By introducing the range information into the network through a dual-stream architecture, we not only improved the relocalization accuracy by about 20% compared with the state-of-the-art deep learning method for pose regression, but also greatly enhanced the system robustness in challenging scenes such as large-scale, dynamic, fast movement, and night-time environments. To the best of our knowledge, this is the first work to solve the indoor relocalization problems based on deep CNNs with RGB-D camera. The method is first evaluated on the Microsoft 7-Scenes data set to show its advantage in accuracy compared with other CNNs. Large-scale indoor relocalization is further presented using our method. The experimental results show that 0.3 m in position and 4° in orientation accuracy could be obtained. Finally, this method is evaluated on challenging indoor data sets collected from motion capture system. The results show that the relocalization performance is hardly affected by dynamic objects, motion blur, or night-time environments.

Visual–inertial odometry (VIO) is widely studied and used in autonomous robots. This article proposes a novel tightly coupled monocular VIO system based on point-line constraints (PLC-VIO). In the front end, PLC-VIO presents a line segment extraction and merging algorithm based on the EDLines method and achieves real-time feature tracking based on the geometric constraints between feature points and lines. In the back end, PLC-VIO reconstructs new 3-D landmarks of feature lines through points on the line and optimizes the states by minimizing a cost function that combines the preintegrated inertial measurement unit (IMU) error term together with the point and line reprojection error terms in a sliding window optimization framework. A loop closure module is also integrated, which enables relocalization and drift elimination. The corresponding experimental evaluations are conducted using public datasets to validate the effectiveness and robustness of the proposed system, and the results show that PLC-VIO can achieve good performance when compared with other state-of-the-art systems and, at the same time, with no compromise to real-time performance. Note to Practitioners —Visual–inertial odometry (VIO) can estimate the states of the rigid body (including position, attitude, and velocity) that is widely used in robotic navigation, autonomous driving, virtual reality (VR), and augmented reality (AR). Aiming at the problem of estimating the states of autonomous robots in the GPS-denied environment, this article proposes a novel VIO system based on the point-line constraints (PLC-VIO). PLC-VIO can not only achieve accurate pose estimation for robots due to the introduction of the line features but also make no concession to real-time performance. Furthermore, PLC-VIO can also enrich the texture features of the environment during the 3-D mapping construction. The corresponding experiments are implemented in public datasets to evaluate the effectiveness, efficiency, and robustness of the proposed system. We believe that PLC-VIO can be widely used in robotic navigation and AR/VR fields to provide accurate position and environment information in real time.

With the development of very high resolution satellite image acquisition technology, remote sensing scene classification has become an important and challenging task. In this article, aiming at tackling this task, we propose a hybrid architecture, i.e., aggregated deep Fisher feature (ADFF), which can make full use of deep convolutional features’ rich semantic information and unsupervised encoding's high robustness. Unlike the previous methods, we first explore the optimal encoding layer in the pretraining CNN model, which naturally fuses the local and global image information in a novel way, making the ability of semantic acquisition further enhanced. ADFF can learn more suitable internal features from the remote sensing data, boosting the final performance. We evaluate our algorithm based on several public datasets, and the results show that our approach achieves superior performance compared with the state-of-the-art methods. The proposed ADFF obtains average classification accuracy of 98.81%, 95.21%, 86.01%, and 88.79%, respectively, on the UC Merced Land-Use, RSSCN7, NWPU-RESISC45 (10% for training), and NWPU-RESISC45 (20% for training) datasets.