Ryohei KANKE Masanobu TAKAHASHI
Amodal Instance Segmentation (AIS) aims to segment the regions of both visible and invisible parts of overlapping objects. The mainstream Mask R-CNN-based methods are unsuitable for thin objects with large overlaps because of their object proposal features with bounding boxes for three reasons. First, capturing the entire shapes of overlapping thin objects is difficult. Second, the bounding boxes of close objects are almost identical. Third, a bounding box contains many objects in most cases. In this paper, we propose a box-free AIS method, Seed-to-Mask, for thin objects with large overlaps. The method specifies a target object using a seed and iteratively extends the segmented region. We have achieved better performance in experiments on artificial data consisting only of thin objects.
Min GAO Gaohua CHEN Jiaxin GU Chunmei ZHANG
Wearing a mask correctly is an effective method to prevent respiratory infectious diseases. Correct mask use is a reliable approach for preventing contagious respiratory infections. However, when dealing with mask-wearing in some complex settings, the detection accuracy still needs to be enhanced. The technique for mask-wearing detection based on YOLOv7-Tiny is enhanced in this research. Distribution Shifting Convolutions (DSConv) based on YOLOv7-tiny are used instead of the 3×3 convolution in the original model to simplify computation and increase detection precision. To decrease the loss of coordinate regression and enhance the detection performance, we adopt the loss function Intersection over Union with Minimum Points Distance (MPDIoU) instead of Complete Intersection over Union (CIoU) in the original model. The model is introduced with the GSConv and VoVGSCSP modules, recognizing the model’s mobility. The P6 detection layer has been designed to increase detection precision for tiny targets in challenging environments and decrease missed and false positive detection rates. The robustness of the model is increased further by creating and marking a mask-wearing data set in a multi environment that uses Mixup and Mosaic technologies for data augmentation. The efficiency of the model is validated in this research using comparison and ablation experiments on the mask dataset. The results demonstrate that when compared to YOLOv7-tiny, the precision of the enhanced detection algorithm is improved by 5.4%, Recall by 1.8%, mAP@.5 by 3%, mAP@.5:.95 by 1.7%, while the FLOPs is decreased by 8.5G. Therefore, the improved detection algorithm realizes more real-time and accurate mask-wearing detection tasks.
Lei WANG Shanmin YANG Jianwei ZHANG Song GU
Human action recognition (HAR) exhibits limited accuracy in video surveillance due to the 2D information captured with monocular cameras. To address the problem, a depth estimation-based human skeleton action recognition method (SARDE) is proposed in this study, with the aim of transforming 2D human action data into 3D format to dig hidden action clues in the 2D data. SARDE comprises two tasks, i.e., human skeleton action recognition and monocular depth estimation. The two tasks are integrated in a multi-task manner in end-to-end training to comprehensively utilize the correlation between action recognition and depth estimation by sharing parameters to learn the depth features effectively for human action recognition. In this study, graph-structured networks with inception blocks and skip connections are investigated for depth estimation. The experimental results verify the effectiveness and superiority of the proposed method in skeleton action recognition that the method reaches state-of-the-art on the datasets.
Highly conflicting evidence that may lead to the counter-intuitive results is one of the challenges for information fusion in Dempster-Shafer evidence theory. To deal with this issue, evidence conflict is investigated based on belief divergence measuring the discrepancy between evidence. In this paper, the pignistic probability transform belief χ2 divergence, named as BBχ2 divergence, is proposed. By introducing the pignistic probability transform, the proposed BBχ2 divergence can accurately quantify the difference between evidence with the consideration of multi-element sets. Compared with a few belief divergences, the novel divergence has more precision. Based on this advantageous divergence, a new multi-source information fusion method is devised. The proposed method considers both credibility weights and information volume weights to determine the overall weight of each evidence. Eventually, the proposed method is applied in target recognition and fault diagnosis, in which comparative analysis indicates that the proposed method can realize the highest accuracy for managing evidence conflict.
Beibei LI Xun RAN Yiran LIU Wensheng LI Qingling DUAN
Fish skin color detection plays a critical role in aquaculture. However, challenges arise from image color cast and the limited dataset, impacting the accuracy of the skin color detection process. To address these issues, we proposed a novel fish skin color detection method, termed VH-YOLOv5s. Specifically, we constructed a dataset for fish skin color detection to tackle the limitation posed by the scarcity of available datasets. Additionally, we proposed a Variance Gray World Algorithm (VGWA) to correct the image color cast. Moreover, the designed Hybrid Spatial Pyramid Pooling (HSPP) module effectively performs multi-scale feature fusion, thereby enhancing the feature representation capability. Extensive experiments have demonstrated that VH-YOLOv5s achieves excellent detection results on the Plectropomus leopardus skin color dataset, with a precision of 91.7%, recall of 90.1%, mAP@0.5 of 95.2%, and mAP@0.5:0.95 of 57.5%. When compared to other models such as Centernet, AutoAssign, and YOLOX-s, VH-YOLOv5s exhibits superior detection performance, surpassing them by 2.5%, 1.8%, and 1.7%, respectively. Furthermore, our model can be deployed directly on mobile phones, making it highly suitable for practical applications.
Xiao’an BAO Shifan ZHOU Biao WU Xiaomei TU Yuting JIN Qingqi ZHANG Na ZHANG
With the popularization of software defined networks, switch migration as an important network management strategy has attracted increasing attention. Most existing switch migration strategies only consider local conditions and simple load thresholds, without fully considering the overall optimization and dynamics of the network. Therefore, this article proposes a switch migration algorithm based on global optimization. This algorithm adds a load prediction module to the migration model, determines the migration controller, and uses an improved whale optimization algorithm to determine the target controller and its surrounding controller set. Based on the load status of the controller and the traffic priority of the switch to be migrated, the optimal migration switch set is determined. The experimental results show that compared to existing schemes, the algorithm proposed in this paper improves the average flow processing efficiency by 15% to 40%, reduces switch migration times, and enhances the security of the controller.
Risheng QIN Hua KUANG He JIANG Hui YU Hong LI Zhuan LI
This paper proposes a determination method of the cascaded number for lumped parameter models (LPMs) of the transmission lines. The LPM is used to simulate long-distance transmission lines, and the cascaded number significantly impacts the simulation results. Currently, there is a lack of a system-level determination method of the cascaded number for LPMs. Based on the theoretical analysis and eigenvalue decomposition of network matrix, this paper discusses the error in resonance characteristics between distributed parameter model and LPMs. Moreover, it is deduced that optimal cascaded numbers of the cascaded π-type and T-type LPMs are the same, and the Γ-type LPM has a lowest analog accuracy. The principle that the maximum simulation frequency is less than the first resonance frequency of each segment is presented. According to the principle, optimal cascaded numbers of cascaded π-type, T-type, and Γ-type LPMs are obtained. The effectiveness of the proposed determination method is verified by simulation.
Ryuichi NAKAJIMA Takafumi ITO Shotaro SUGITANI Tomoya KII Mitsunori EBARA Jun FURUTA Kazutoshi KOBAYASHI Mathieu LOUVAT Francois JACQUET Jean-Christophe ELOY Olivier MONTFORT Lionel JURE Vincent HUARD
We evaluated soft-error tolerance by heavy-ion irradiation test on three-types of flip-flops (FFs) named the standard FF (STDFF), the dual feedback recovery FF (DFRFF), and the DFRFF with long delay (DFRFFLD) in 22 and 65 nm fully-depleted silicon on insulator (FD-SOI) technologies. The guard-gate (GG) structure in DFRFF mitigates soft errors. A single event transient (SET) pulse is removed by the C-element with the signal delayed by the GG structure. DFRFFLD increases the GG delay by adding two more inverters as delay elements. We investigated the effectiveness of the GG structure in 22 and 65 nm. In 22 nm, Kr (40.3 MeV-cm2/mg) and Xe (67.2 MeV-cm2/mg) irradiation tests revealed that DFRFFLD has sufficient soft-error tolerance in outer space. In 65 nm, the relationship between GG delay and CS reveals the GG delay time which no error was observed under Kr irradiation.
This letter deals with joint carrier frequency offset (CFO) and direction of arrival (DOA) estimation based on the minimum variance distortionless response (MVDR) criterion for interleaved orthogonal frequency division multiple access (OFDMA)/space division multiple access (SDMA) uplink systems. In order to reduce the computational load of two-dimensional searching based methods, the proposed method includes only once polynomial CFO rooting and does not require DOA paring, hence it raises the searching efficiency. Several simulation results are provided to illustrate the effectiveness of the proposed method.
Ze Fu GAO Wen Ge YANG Yi Wen JIAO
Space is becoming increasingly congested and contested, which calls for effective means to conduct effective monitoring of high-value space assets, especially in Space Situational Awareness (SSA) missions, while there are imperfections in existing methods and corresponding algorithms. To overcome such a problem, this letter proposes an algorithm for accurate Connected Element Interferometry (CEI) in SSA based on more interpolation information and iterations. Simulation results show that: (i) after iterations, the estimated asymptotic variance of the proposed method can basically achieve uniform convergence, and the ratio of it to ACRB is 1.00235 in δ0 ∈ [-0.5, 0.5], which is closer to 1 than the current best AM algorithms; (ii) In the interval of SNR ∈ [-14dB, 0dB], the estimation error of the proposed algorithm decreases significantly, which is basically comparable to CRLB (maintains at 1.236 times). The research of this letter could play a significant role in effective monitoring and high-precision tracking and measurement with significant space targets during futuristic SSA missions.
Longye WANG Houshan LIU Xiaoli ZENG Qingping YU
This letter presented several new constructions of complementary sets (CSs) with flexible sequence lengths using matrix transformations. The constructed CSs of size 4 have different lengths, namely N + L and 2N + L, where N and L are the lengths for which complementary pairs exist. Also, presented CSs of size 8 have lengths N + P, P + Q and 2P + Q, where N is length of complementary pairs, P and Q are lengths of CSs of size 4 exist. The achieved designs can be easily extended to a set size of 2n+2 by recursive method. The proposed constructions generalize some previously reported constructions along with generating CSs under fewer constraints.
Chao HE Xiaoqiong RAN Rong LUO
Cyclic codes are a subclass of linear codes and have applications in consumer electronics, data storage systems, and communication systems as they have efficient encoding and decoding algorithms. Let C(t,e) denote the cyclic code with two nonzero αt and αe, where α is a generator of 𝔽*3m. In this letter, we investigate the ternary cyclic codes with parameters [3m - 1, 3m - 1 - 2m, 4] based on some results proposed by Ding and Helleseth in 2013. Two new classes of optimal ternary cyclic codes C(t,e) are presented by choosing the proper t and e and determining the solutions of certain equations over 𝔽3m.
Changhui CHEN Haibin KAN Jie PENG Li WANG
Permutation polynomials have important applications in cryptography, coding theory and combinatorial designs. In this letter, we construct four classes of permutation polynomials over 𝔽2n × 𝔽2n, where 𝔽2n is the finite field with 2n elements.
Yingzhong ZHANG Xiaoni DU Wengang JIN Xingbin QIAO
Boolean functions with a few Walsh spectral values have important applications in sequence ciphers and coding theory. In this paper, we first construct a class of Boolean functions with at most five-valued Walsh spectra by using the secondary construction of Boolean functions, in particular, plateaued functions are included. Then, we construct three classes of Boolean functions with five-valued Walsh spectra using Kasami functions and investigate the Walsh spectrum distributions of the new functions. Finally, three classes of minimal linear codes with five-weights are obtained, which can be used to design secret sharing scheme with good access structures.
Longye WANG Chunlin CHEN Xiaoli ZENG Houshan LIU Lingguo KONG Qingping YU Qingsong WANG
Spatial modulation (SM) is a type of multiple-input multiple-output (MIMO) technology that provides several benefits over traditional MIMO systems. SM-MIMO is characterized by its unique transmission principle, which results in lower costs, enhanced spectrum utilization, and reduced inter-channel interference. To optimize channel estimation performance over frequency-selective channels in the spatial modulation system, cross Z-complementary pairs (CZCPs) have been proposed as training sequences. The zero correlation zone (ZCZ) properties of CZCPs for auto-correlation sums and cross-correlation sums enable them to achieve optimal channel estimation performance. In this paper, we systematically construct CZCPs based on binary Golay complementary pairs and binary Golay complementary pairs via Turyn’s method. We employ a special matrix operation and concatenation method to obtain CZCPs with new lengths 2M + N and 2(M + L), where M and L are the lengths of binary GCP, and N is the length of binary GCP via Turyn’s method. Further, we obtain the perfect CZCP with new length 4N and extend the lengths of CZCPs.
Yuto ARIMURA Shigeru YAMASHITA
Stochastic Computing (SC) allows additions and multiplications to be realized with lower power than the conventional binary operations if we admit some errors. However, for many complex functions which cannot be realized by only additions and multiplications, we do not know a generic efficient method to calculate a function by using an SC circuit; it is necessary to realize an SC circuit by using a generic method such as polynomial approximation methods for such a function, which may lose the advantage of SC. Thus, there have been many researches to consider efficient SC realization for specific functions; an efficient SC square root circuit with a feedback circuit was proposed by D. Wu et al. recently. This paper generalizes the SC square root circuit with a feedback circuit; we identify a situation when we can implement a function efficiently by an SC circuit with a feedback circuit. As examples of our generalization, we propose SC circuits to calculate the n-th root calculation and division. We also show our analysis on the accuracy of our SC circuits and the hardware costs; our results show the effectiveness of our method compared to the conventional SC designs; our framework may be able to implement a SC circuit that is better than the existing methods in terms of the hardware cost or the calculation error.
Chengyu WU Jiangshan QIN Xiangyang LI Ao ZHAN Zhengqiang WANG
Real-time matting is a challenging research in deep learning. Conventional CNN (Convolutional Neural Networks) approaches are easy to misjudge the foreground and background semantic and have blurry matting edges, which result from CNN’s limited concentration on global context due to receptive field. We propose a real-time matting approach called RMViT (Real-time matting with Vision Transformer) with Transformer structure, attention and content-aware guidance to solve issues above. The semantic accuracy improves a lot due to the establishment of global context and long-range pixel information. The experiments show our approach exceeds a 30% reduction in error metrics compared with existing real-time matting approaches.
Yuto HOSHINO Hiroki KAWAKAMI Hiroki MATSUTANI
Federated learning is a distributed machine learning approach in which clients train models locally with their own data and upload them to a server so that their trained results are shared between them without uploading raw data to the server. There are some challenges in federated learning, such as communication size reduction and client heterogeneity. The former can mitigate the communication overheads, and the latter can allow the clients to choose proper models depending on their available compute resources. To address these challenges, in this paper, we utilize Neural ODE based models for federated learning. The proposed flexible federated learning approach can reduce the communication size while aggregating models with different iteration counts or depths. Our contribution is that we experimentally demonstrate that the proposed federated learning can aggregate models with different iteration counts or depths. It is compared with a different federated learning approach in terms of the accuracy. Furthermore, we show that our approach can reduce communication size by up to 89.4% compared with a baseline ResNet model using CIFAR-10 dataset.
Haochen LYU Jianjun LI Yin YE Chin-Chen CHANG
The purpose of Facial Beauty Prediction (FBP) is to automatically assess facial attractiveness based on human aesthetics. Most neural network-based prediction methods do not consider the ranking information in the task. For scoring tasks like facial beauty prediction, there is abundant ranking information both between images and within images. Reasonable utilization of these information during training can greatly improve the performance of the model. In this paper, we propose a novel end-to-end Convolutional Neural Network (CNN) model based on ranking information of images, incorporating a Rank Module and an Adaptive Weight Module. We also design pairwise ranking loss functions to fully leverage the ranking information of images. Considering training efficiency and model inference capability, we choose ResNet-50 as the backbone network. We conduct experiments on the SCUT-FBP5500 dataset and the results show that our model achieves a new state-of-the-art performance. Furthermore, ablation experiments show that our approach greatly contributes to improving the model performance. Finally, the Rank Module with the corresponding ranking loss is plug-and-play and can be extended to any CNN model and any task with ranking information. Code is available at https://github.com/nehcoah/Rank-Info-Net.
In the software testing phase, software reliability growth models (SRGMs) are commonly used to evaluate the reliability of software systems. Traditional SRGMs are restricted by their assumption of a continuous growth pattern for the failure detection rate (FDR) throughout the testing phase. However, the assumption is compromised by Change-Point phenomena, where FDR fluctuations stem from variations in testing personnel or procedural modifications, leading to reduced prediction accuracy and compromised software reliability assessments. Therefore, the objective of this study is to improve software reliability prediction using a novel approach that combines genetic algorithm (GA) and deep learning-based SRGMs to account for the Change-point phenomenon. The proposed approach uses a GA to dynamically combine activation functions from various deep learning-based SRGMs into a new mutated SRGM called MuSRGM. The MuSRGM captures the advantages of both concave and S-shaped SRGMs and is better suited to capture the change-point phenomenon during testing and more accurately reflect actual testing situations. Additionally, failure data is treated as a time series and analyzed using a combination of Long Short-Term Memory (LSTM) and Attention mechanisms. To assess the performance of MuSRGM, we conducted experiments on three distinct failure datasets. The results indicate that MuSRGM outperformed the baseline method, exhibiting low prediction error (MSE) on all three datasets. Furthermore, MuSRGM demonstrated remarkable generalization ability on these datasets, remaining unaffected by uneven data distribution. Therefore, MuSRGM represents a highly promising advanced solution that can provide increased accuracy and applicability for software reliability assessment during the testing phase.