Semi-supervised learning aims to leverage numerous unlabeled data to improve the model performance. Current semi-supervised 3D object detection methods typically use a teacher to generate pseudo labels for a student, and the quality of the pseudo labels is essential for the final performance. In this paper, we propose PatchTeacher, which focuses on partial scene 3D object detection to provide high-quality pseudo labels for the student. Specifically, we divide a complete scene into a series of patches and feed them to our PatchTeacher sequentially. PatchTeacher leverages the low memory consumption advantage of partial scene detection to process point clouds with a high-resolution voxelization, which can minimize the information loss of quantization and extract more fine-grained features. However, it is non-trivial to train a detector on fractions of the scene. Therefore, we introduce three key techniques, i.e., Patch Normalizer, Quadrant Align, and Fovea Selection, to improve the performance of PatchTeacher. Moreover, we devise PillarMix, a strong data augmentation strategy that mixes truncated pillars from different LiDAR scans to generate diverse training samples and thus help the model learn more general representation. Extensive experiments conducted on Waymo and ONCE datasets verify the effectiveness and superiority of our method and we achieve new state-of-the-art results, surpassing existing methods by a large margin. Codes are available at https://github.com/LittlePey/PTPM.

Monocular 3D detection is a challenging task due to the lack of accurate 3D information. Existing approaches typically rely on geometry constraints and dense depth esti-mates to facilitate the learning, but often fail to fully ex-ploit the benefits of three-dimensional feature extraction in frustum and 3D space. In this paper, we propose Occu- pancyM3D, a method of learning occupancy for monocu-lar 3D detection. It directly learns occupancy in frustum and 3D space, leading to more discriminative and informative 3D features and representations. Specifically, by using synchronized raw sparse LiDAR point clouds, we define the space status and generate voxel-based occupancy labels. We formulate occupancy prediction as a simple classification problem and design associated occupancy losses. Re-sulting occupancy estimates are employed to enhance orig-inal frustum/3D features. As a result, experiments on KITTI and Waymo open datasets demonstrate that the proposed method achieves a new state of the art and surpasses other methods by a significant margin.

Training deep models for LiDAR semantic segmentation is challenging due to the inherent sparsity of point clouds. Utilizing temporal data is a natural remedy against the spar-sity problem as it makes the input signal denser. However, previous multi-frame fusion algorithms fall short in utilizing sufficient temporal information due to the memory constraint, and they also ignore the informative temporal images. To fully exploit rich information hidden in long-term temporal point clouds and images, we present the Temporal Aggre-gation Network, termed TASeg. Specifically, we propose a Temporal LiDAR Aggregation and Distillation (TLAD) algorithm, which leverages historical priors to assign dif-ferent aggregation steps for different classes. It can largely reduce memory and time overhead while achieving higher accuracy. Besides, TLAD trains a teacher injected with gt priors to distill the model, further boosting the performance. To make full use of temporal images, we design a Temporal Image Aggregation and Fusion (TIAF) module, which can greatly expand the camera FOVand enhance the present features. Temporal LiDAR points in the camera FOV are used as mediums to transform temporal image features to the present coordinate for temporal multi-modal fusion. Moreover, we develop a Static-Moving Switch Augmentation (SMSA) algorithm, which utilizes sufficient temporal information to enable objects to switch their motion states freely, thus greatly increasing static and moving training samples. Our TASeg ranks 1st††the date of CVPR deadline, i.e., 2023-11-18 07:59 AM UTC. on three challenging tracks, i.e., SemanticKITTI single-scan track, multi-scan track and nuScenes LiDAR segmentation track, strongly demonstrating the superiority of our method. Codes are available at https://github.com/LittlePey/TASeg.

Currently, there is a growing global concern regarding data privacy and security, particularly in the field of personal identification and authentication. Traditional biometric identification technologies are highly favored for their ease of use and high accuracy, but they fall short in ensuring liveness detection, making them susceptible to deception and forgery threats. This study focuses on personal identification and authentication based on a multi-task electroencephalogram (EEG) database, proposing an innovative model framework for these purposes. To validate the effectiveness of this model, we established a multitask EEG database containing data from 24 subjects engaged in five mental tasks. Each subject underwent four sessions, with each session consisting of 125 trials, and session intervals ranging from days to months. In this framework, we employed the EEGNet model for personal identification. It directly utilized preprocessed EEG data as input, mapping input signals to a new embedding space to extract identity features, ultimately achieving accurate individual personal identification. For the personal authentication module, we proposed the SiamEEGNet model, combining concepts from EEGNet and Siamese networks. This model comprised two EEGNet sub-networks with identical model parameters. Unlike the personal identification module, we removed the classification module from the EEGNet model in the SiamEEGNet model. Instead, we introduced a distance measurement layer to calculate the distance or similarity between the outputs of the two sub-networks, thereby achieving personal authentication. We conducted experiments for both personal identification and authentication. In the identification experiments, the proposed EEGNet model demonstrated outstanding performance with an average recognition accuracy of 99.84%. The model achieved a balance between precision and sensitivity, as reflected in high F1 score values. In the personal authentication experiments, the new SiamEEGNet model achieved a False Rejection Rate (FRR) of 2.29% and a False Acceptance Rate (FAR) of 4.75%. The experimental results collectively indicate the significant effectiveness of the proposed model framework.

The power spectra estimated from the brain recordings are the mixed representation of aperiodic transient activity and periodic oscillations, i.e., aperiodic component (AC) and periodic component (PC). Quantitative neurophysiology requires precise decomposition preceding parameterizing each component. However, the shape, statistical distribution, scale, and mixing mechanism of AC and PCs are unclear, challenging the effectiveness of current popular parametric models such as FOOOF, IRASA, BOSC, etc. Here, ξ- π was proposed to decompose the neural spectra by embedding the nonparametric spectra estimation with penalized Whittle likelihood and the shape language modeling into the expectation maximization framework. ξ- π was validated on the synthesized spectra with loss statistics and on the sleep EEG and the large sample iEEG with evaluation metrics and neurophysiological evidence. Compared to FOOOF, both the simulation presenting shape irregularities and the batch simulation with multiple isolated peaks indicated that ξ- π improved the fit of AC and PCs with less loss and higher F1-score in recognizing the centering frequencies and the number of peaks; the sleep EEG revealed that ξ- π produced more distinguishable AC exponents and improved the sleep state classification accuracy; the iEEG showed that ξ- π approached the clinical findings in peak discovery. Overall, ξ- π offered good performance in the spectra decomposition, which allows flexible parameterization using descriptive statistics or kernel functions. ξ- π is a seminal tool for brain signal decoding in fields such as cognitive neuroscience, brain-computer interface, neurofeedback, and brain diseases.

One-to-one set matching is a key design for DETR to establish its end-to-end capability, so that object detection does not require a hand-crafted NMS (non-maximum suppression) to remove duplicate detections. This end-to-end signature is important for the versatility of DETR, and it has been generalized to broader vision tasks. However, we note that there are few queries assigned as positive samples and the one-to-one set matching significantly reduces the training efficacy of positive samples. We propose a simple yet effective method based on a hybrid matching scheme that combines the original one-to-one matching branch with an auxiliary one-to-many matching branch during training. Our hybrid strategy has been shown to significantly improve accuracy. In inference, only the original one-to-one match branch is used, thus maintaining the end-to-end merit and the same inference efficiency of DETR. The method is namedℋ-DETR, and it shows that a wide range of representative DETR methods can be consistently improved across a wide range of visual tasks, including Deformable-DETR, PETRv2, PETR, and TransTrack, among others. Code is available at: https://github.com/HDETR.

Current LiDAR-only 3D detection methods inevitably suffer from the sparsity of point clouds. Many multi-modal methods are proposed to alleviate this issue, while different representations of images and point clouds make it difficult to fuse them, resulting in suboptimal performance. In this paper, we present a novel multi-modal framework SFD (Sparse Fuse Dense), which utilizes pseudo point clouds generated from depth completion to tackle the issues mentioned above. Different from prior works, we propose a new RoI fusion strategy 3D-GAF (3D Grid-wise Attentive Fusion) to make fuller use of information from different types of point clouds. Specifically, 3D-GAF fuses 3D RoI features from the pair of point clouds in a grid-wise attentive way, which is more fine- grained and more precise. In addition, we propose a SynAugment (Synchronized Augmentation) to enable our multi-modal framework to utilize all data augmentation approaches tailored to LiDAR-only methods. Lastly, we customize an effective and efficient feature extractor CPConv (Color Point Convolution) for pseudo point clouds. It can explore 2D image features and 3D geometric features of pseudo point clouds simultaneously. Our method holds the highest entry on the KITTI car 3D object detection leaderboard††On the date of CVPR deadline, i.e., Nov.16, 2021, demonstrating the effectiveness of our SFD. Code will be made publicly available.

With the rapid development of various computing technologies, the constraints of data processing capabilities gradually disappeared, and more data can be simultaneously processed to obtain better performance compared to conventional methods. As a standard statistical analysis method that has been widely used in many fields, Independent Component Analysis (ICA) provides a new way for motion detection by extracting the foreground without precisely modeling the background. However, most existing ICA-based motion detection algorithms use only two-channel data for source separation and simply generate the observation vectors by decomposing and reconstructing the images by row, hence they cannot obtain an integrated and accurate shape of the moving objects in complex scenes. In this article, we propose a refined ICA algorithm for motion detection (RICA-MD), which fuses a larger number of channels than conventional ICA-based motion detection algorithms to provide more effective information for foreground extraction. Meanwhile, we propose four novel methods for generating observation vectors to further cover the diverse motion styles of the moving objects. These improvements enable RICA-MD to effectively deal with slowly moving objects, which are difficult to detect using conventional methods. Our quantitative evaluation in multiple scenes shows that our proposed method is able to achieve a better performance at an acceptable cost of false alarms.

This paper is focused on the experimental approach to explore the potential of independent component analysis (ICA) in the context of motor imagery (MI)-based brain-computer interface (BCI). We presented a simple and efficient algorithmic framework of ICA-based MI BCI (ICA-MIBCI) for the evaluation of four classical ICA algorithms (Infomax, FastICA, Jade, and Sobi) as well as a simplified Infomax (sInfomax). Two novel performance indexes, self-test accuracy and the number of invalid ICA filters, were employed to assess the performance of MIBCI based on different ICA variants. As a reference method, common spatial pattern (CSP), a commonly-used spatial filtering method, was employed for the comparative study between ICA-MIBCI and CSP-MIBCI. The experimental results showed that sInfomax-based spatial filters exhibited significantly better transferability in session to session and subject to subject transfer as compared to CSP-based spatial filters. The online experiment was also introduced to demonstrate the practicability and feasibility of sInfomax-based MIBCI. However, four classical ICA variants, especially FastICA, Jade, and Sobi, performed much worse as compared to sInfomax and CSP in terms of classification accuracy and stability. We consider that conventional ICA-based spatial filtering methods tend to be overfitting while applied to real-life electroencephalogram data. Nevertheless, the sInfomax-based experimental results indicate that ICA methods have a great space for improvement in the application of MIBCI. We believe that this paper could bring forth new ideas for the practical implementation of ICA-MIBCI.

With the installation of enormous public safety and transportation infrastructure cameras, video analytics has come to play an essential part in public safety. Typically, video analytics is to collectively leverage the advanced computer vision (CV) and artificial intelligence (AI) to solve the four-W problem. That is to identify Who has done something (What) at a specific place (Where) at some time (When). According to the difference of latency requirements, video analytics can be applied to postevent retrospective analysis, such as archive management, search, forensic investigation and real-time live video stream analysis, such as situation awareness, alerting, and interested object (criminal suspect/missing vehicle) detection. The latter is characterized as having higher requirements on hardware resources as the sophisticated image processing algorithms under the hood. However, analyzing large-scale live video streams on the Cloud is impractical as the edge solution that conducts the video analytics on (or close to) the camera provides a silvering light. Analyzing live video streams on the edge is not trivial due to the constrained hardware resources on edge. The AI-dominated video analytics requires higher bandwidth, consumes considerable CPU/GPU resources for processing, and demands larger memory for caching. In this paper, we review the applications, algorithms, and solutions that have been proposed recently to facilitate edge video analytics for public safety.

In this paper, we envision the future connected and autonomous vehicles (CAVs) as a sophisticated computer on wheels, with substantial on-board sensors as data sources and a variety of services running on top to support autonomous driving or other functions. In general, these services are computationally expensive, especially for the machine learning based applications (e.g., CNN-based object detection). Nevertheless, the on-board computation unit possess limited compute resources, raising a huge challenge to deploy these computation-intensive services on the vehicle. On the contrary, the cloud-based architecture conceptually with unconstrained resources suffers from unexpected extended latency that attributes to the large-scale Internet data transmission; thus, adversely affecting the services' real-time performance, quality of services and user experiences. To address this dilemma, inspired by the promising edge computing paradigm, we propose to build an Open Vehicular Data Analytics Platform (OpenVDAP) for CAVs, which is a full-stack edge based platform including an on-board computing/communication unit, an isolation-supported and security & privacy-preserved vehicle operation system, an edge-aware application library, as well as an optimal workload of?oading and scheduling strategy, allowing CAVs to dynamically detect each service's status, computation overhead and the optimal of?oading destination so that each service could be finished within an acceptable latency and limited bandwidth consumption. Most importantly, contrast to the proprietary platform, OpenVDAP is an open-source platform that offers free APIs and real-?eld vehicle data to the researchers and developers in the community, allowing them to deploy and evaluate applications on the real environment.

Accurate maps are increasingly important with the growth of smart phones and the development of location-based services. Several crowdsourcing based map generation protocols that rely on users to provide their traces have been proposed. Being creative, however, those methods pose a significant threat to user privacy as the traces can easily imply user behavior patterns. On the flip side, crowdsourcing-based map generation method does need individual locations. To address the issue, we present a systematic participatory-sensing-based high-quality map generation scheme, PMG, that meets the privacy demand of individual users. To be specific, the individual users merely need to upload unorganized sparse location points to reduce the risk of exposing users’ traces and utilize theCrust, a technique from computational geometry for curve reconstruction, to estimate the unobserved map as well as evaluate the degree of privacy leakage. Experiments show that our solution is able to generate high-quality maps for a real environment that is robust to noisy data. The difference between the ground-truth map and the produced map is less than 10 m, even when the collected locations are about 32 m apart after clustering for the purpose of removing noise.

On-air access of individual sensor node (called remote control) is an indispensable function in operational wireless sensor networks, for purposes like network management and real-time information delivery. To realize reliable and efficient remote control in a wireless sensor network (WSN), however, is extremely challenging, due to the stringent resource constraints and intrinsically unrealizable wireless communication. In this paper, we propose TeleAdjusting, a ready-to-use protocol to remotely control any individual node in a WSN. We develop a coding scheme for addressing on the cost-optimal reverse routing tree. In the address of each node, all its upstream relaying nodes are implicitly encoded. Then through a distributed prefix matching process between the local address and the destination address, a packet used for remote control is forwarded along a cost-optimal path. Moreover, TeleAdjusting incorporates opportunistic forwarding into the addressing process, so as to improve the network performance in terms of reliability and energy efficiency. We implement TeleAdjusting with TinyOS and evaluate its performance through extensive simulations and experiments. The results demonstrate that compared to the existing protocols, TeleAdjusting can provide high performance of remote control, which is as reliable as network-wide flooding and much more efficient than remote control through a pre-determined path.

The energy efficiency and delivery robustness are two critical issues for low duty cycled wireless sensor networks. The asynchronous receiver-initiated duty cycling media access control (MAC) protocols have shown the effectiveness through various studies. In receiver-initiated MACs, packet transmission is triggered by the probe of receiver. However, it suffers from the performance degradation incurred by packet collision, especially under bursty traffic. Several protocols have been proposed to address this problem, but their performance is restricted by the unnecessary backoff time and long negotiation process. In this paper, we present Contention Detectable MAC (CD-MAC), an energy efficient and robust duty-cycled MAC for general wireless sensor network applications. By exploring the temporal diversity of the acknowledgements, a receiver recognizes the potential senders and subsequently polls individual senders one by one. We further design efficient algorithm to avoid the possible acknowledgement collision. We implement CD-MAC in TinyOS and evaluate the performance on an indoor testbed with single-hop and multi-hop networks. The results show that CD-MAC can significantly improve throughput by 1.72 times compared with the state-of-the-art receiver-initiated MAC protocol under bursty traffic loads. The results also demonstrate that CD-MAC can effectively mitigate the influence of hidden terminal problem and adapt to network dynamics well.

Smartphones nowadays are installed with diverse applications, each of which consumes energy and bandwidth. As more and more applications are crowded into a smart- phone, they cause serious problems with regard to battery life and bandwidth utilization. Existing proposals to tackle such challenges usually resort to two ways: avoiding energy- consuming network activities or improving communication efficiency in terms of power consumption. Those approaches either affect the smartphone users' experience, or offer little benefit in prolonging the battery life. Motivated by insightful understanding of users' habit, we in this paper propose a novel approach to orchestrate network activities of smartphone applications, based on user's habit. We implement our approach on smartphones as a middleware service called NetMaster. The performance evaluation with real traces shows that NetMaster reduces energy consumption of network activities by 77.8% in average and increases network bandwidth utilization by over 200%. The user experience is surprisingly well preserved. The chance of undesired interrupt during normal usage is less than 1%.

Accurate maps are increasingly important with the growth of smart phones and the development of location-based services. Several crowdsourcing based map generation protocols have been proposed that rely on volunteers to provide their traces. Being creative, however, those methods pose a significant threat to user privacy as the traces can easily imply user behavior patterns. On the flip side, crowdsourcing-based map generation method does need individual locations. To address the issue, we present a systematic participatory-sensing-based high-quality map generation scheme, PMG, that meets the privacy demand of individual users. In this approach, individual users merely need to upload unorganized sparse location points so as to reduce the risk of exposing privacy, while the server generates accurate maps with unorganized points, instead of user traces. Experiments show that our solution is able to generate high-quality maps for a real environment that is robust to noisy data. The difference between the ground-truth map and the produced map is <; 10m, even when the collected locations are about 32m apart after clustering for the purpose of removing noise.

Collision resolution is a crucial issue in wireless sensor networks. The existing approaches of collision resolution have drawbacks with respect to energy efficiency and processing latency. In this paper, we propose ST AIRS, a time and energy efficient collision resolution mechanism for wireless sensor networks. STAIRS incorporates the constructive interference technique in its design and explicitly forms superimposed colliding signals. Through extensive observations and theoretical analysis, we show that the RSSI of the superimposed signals exhibit stairs-like phenomenon with different number of contenders. That principle offers an attractive feature to efficiently distinguish multiple contenders and in turn makes collision-free schedules for channel access. In the design and implementation of STAIRS, we address practical challenges such as contenders alignment, online detection of RSSI change points, and fast channel assignment. The experiments on real testbed show that STARIS realizes fast and effective collision resolution, which significantly improves the network performance in terms of both latency and throughput.

We consider the problem of monitoring soil moisture evolution using a wireless network of in-situ underground sensors. To reduce cost and prolong lifetime, it is highly desirable to rely on fewer measurements and estimate with higher accuracy the original signal (the temporal evolution of soil moisture). In this article, we explore the use of results from the theory of sparse sampling, including Compressive Sensing (CS) and Matrix Completion (MC), in this application context. We first consider the problem of reconstructing the soil moisture process at a single location using CS. Our physical constraint leads to very sparse measurement matrices, which makes finding a suitable representation basis very challenging: it needs to make the underlying signal sufficiently sparse while at the same time being sufficiently incoherent with the measurement matrix, two common preconditions for CS techniques to work well. We construct a representation basis by exploiting unique features of soil moisture evolution and show that this basis attains a very good tradeoff between its ability to sparsify the signal and its incoherence with measurement matrices that are consistent with our physical constraints. We next consider the problem of jointly reconstructing soil moisture processes at multiple locations, assuming sparse measurements can be taken at each location. We show that the spatial soil moisture process enjoys a low-rank property, a priority for MC. Accordingly, we introduce a spatiotemporal measurement matrix and apply the MC framework to reconstruct the soil moisture field. Extensive numerical evaluation is performed on both real, high-resolution soil moisture data and simulated data and through comparison with a closed-loop scheduling approach. Our results demonstrate that, for a single location, a uniform measurement scheduling followed by CS recovery results in a very nice tradeoff between estimation accuracy, sampling rate, flexibility, and feasibility in implementation. When multiple locations are available, our results show that joint reconstruction using MC in general produces better estimation accuracy than using a single location alone, but it requires the use of independent and random measurement schedules across locations. We also show that these sparse sampling techniques can be augmented so as to be robust against sporadic data outliers/corruption caused by, for example, intermittent sensor faults.

We consider the problem of monitoring soil moisture evolution using a wireless network of in-situ underground sensors. To reduce cost and prolong lifetime, it is highly desirable to rely on fewer measurements and estimate with higher accuracy the original signal (soil moisture temporal evolution). In this paper we explore results from the compressive sensing (CS) literature and examine their applicability to this problem. Our main challenge lies in the selection of two matrices, the measurement matrix and a representation basis. The physical constraints of our problem make it highly non-trivial to select these matrices, so that the latter can sufficient sparsify the underlying signal while at the same time be sufficiently incoherent with the former, two common pre-conditions for CS techniques to work well. We construct a representation basis by exploiting unique features of soil moisture evolution. We show that this basis attains very good tradeoff between its ability to sparsify the signal and its incoherence with measurement matrices that are consistent with our physical constraints. Extensive numerical evaluation is performed on both real, high-resolution soil moisture data and simulated data, and through comparison with a closed-loop scheduling approach. Our results demonstrate that our approach is extremely effective in reconstructing the soil moisture process with high accuracy and low sampling rate.

We study the problem of optimal sensor placement in the context of soil moisture sensing. We show that the soil moisture data possesses some unique features that can be used together with the commonly used Gaussian assumption to construct more scalable, robust, and better performing placement algorithms. Specifically, there exists a coarse-grained monotonic ordering of locations in their soil moisture level over time, both in terms of its first and second moments, a feature much more stable than the soil moisture process itself at these locations. This motivates a clustered sensor placement scheme, where locations are classified into clusters based on the ordering of the mean, with the number of sensors placed in each cluster determined by the ordering of the variances. We show that under idealized conditions the greedy mutual information maximization algorithm applied globally is equivalent to that applied cluster by cluster, but the latter has the advantage of being more scalable. Extensive numerical experiments are performed on a set of three-dimensional soil moisture data generated by a state-of-the-art soil moisture simulator. Our results show that our clustering approach outperforms applying the same algorithms globally, and is very robust to lack of training and errors in training data.

The paper investigates the possibility of using empirical mode decomposition (EMD) method to detect the mu rhythm of motor imagery EEG signal. Recently the mu rhythm by motor imagination has been used as a reliable EEG pattern for brain-computer interface (BCI) system. Considering the non-stationary characteristics of the motor imagery EEG, the EMD method is proposed to detect the mu rhythm during left and right hand movement imagination. By analyzing the instantaneous amplitude and instantaneous frequency of the intrinsic mode functions (IMFs), the mu rhythm can be detected. And by Hilbert marginal spectrum, the ERD/ERS phenomenon of mu rhythm can be found. The results in this paper demonstrate that the EMD method is a effective time-frequency analysis tool for non-stationary EEG signal.