Search

Progressive training has unfolded its superiority on a wide range of downstream tasks. However, it may fail in fine-grained recognition (FGR) due to special challenges with high intra-class and low inter-class variances. In this paper, we propose an active self-pace learning method to exploit the full potential of progressive training strategy in FGR. The key innovation of our design is to integrate submodular optimization and self-pace learning into a maximum–minimum optimization framework. The submodular optimization is regarded as a dynamic regularization to select active sample groups in each training round for restricting the search space of self-pace optimization. This can overcome the limitation of traditional self-pace learning that is easily trapped into local minimums when facing challenging samples. Extensive experiments on three public FGR datasets show that the proposed method can win at least 1.5% performance gain in various kinds of network backbones including swin-transformer. • We are the first to exploit the sub–modularity for active sample selection. By our problem formulation, the optimal category subsets can be progressively selected for obtaining steady cumulative gain. • We combine submodular optimization with self-paced learning to generate a collaborated maximum–minimum optimization framework. The constructed framework can achieve smooth and stable progressive learning through using active samples to restrict the search space of self-paced optimization. • The proposed collaborated optimization framework can be deployed on various types of FGR networks. Extensive experiments on three fine-grained recognition datasets can verify the superiority of progressive training, where the averaged recognition gain surpasses 1.5%. [ABSTRACT FROM AUTHOR]

Estimating human pose from images is the key to enabling machines to understand human actions. Existing works on human pose estimation mainly focus on designing more resultful deep neural networks to regress the locations of human joints. Although the human pose is obedient to anatomy and shows rich anatomical features, reasoning human body structure by machine in a complex environment is still an open problem. This paper proposes AnatPose which can effectively capture the structural dependency among human body parts by both deep neural network architecture and learning objectives: (1) For the deep neural network architecture, a bidirectional learning paradigm is proposed to learn body-part proportions and dependencies by organizing human body parts as sequential data. This innovation enables the messages to pass in a bidirectional way and makes the human body exchange information about each part deeper during training. (2) For the learning objective, the proposed AnatPose learns a probabilistic representation of multi-scale anatomical features, including keypoint heatmaps, bone heatmaps, and symmetry heatmaps. This innovation enables the multi-scale anatomical features to successfully capture the structural dependency at both low-level joints and high-level associations from the anatomical priors of the human body. Extensive experimental results demonstrate that the proposed AnatPose shows state-of-the-art performance on three challenging datasets. It achieves a PCK@0.2 detection rate of 95.2% on the LSP dataset, a PCKh@0.5 detection rate of 92.9% on the MPII dataset, and an mAP of 76.6% on the Microsoft COCO dataset. Benefiting from its state-of-the-art accuracy, the proposed approach is expected to be widely used in various human pose estimation-driven applications. • AnatPose organizes human body parts as anatomically sequential data. • AnatPose learns the anatomical properties by a bidirectional recurrent neural network. • An anatomically significant objective is designed for optimizing the neural network. • Extensive results show superior performance on human pose estimation. [ABSTRACT FROM AUTHOR]

3D human pose estimation plays an important role in various human-machine interactive applications, but how to effectively extract and represent the kinematical features of human body structure in video has always been a challenge. This paper presents some inspiring observations on the human body properties that hold heuristic patterns of human poses: 1) There is distinct temporal coherence in any kind of human pose; 2) there exist evident spatial and temporal correlations among local joints even though the human is doing complex actions. According to the observed patterns, a locally structured feature encoder and a spatial–temporal feature transform are proposed for kinematics-aware feature extraction and enhancement. Unlike existing works directly projecting every bone joint to pose features without distinction, the proposed locally-structured feature encoder maps the local connection property of human body structure to kinematical features which are neural embeddings extracted from both local and global groups of human bone joints. Since the local and global bone-joint groups are pre-defined according to human body kinematics, the kinematical features are able to represent body kinematics. The kinematical features are then transformed by the proposed spatial–temporal feature transform to enhance the spatial and temporal correlations among human bone joints. The overall framework well promotes the representation of human body kinematics for 3D pose estimation. Extensive experimental results on commonly used datasets show that the mean per joint position error (MPJPE) is significantly reduced when compared with state-of-the-art methods under the same experimental condition. The improvement is expected to promote machines to better understand human poses for building superior human-centered automation systems. • Spatial–temporal kinematic-awareness is studied for 3D human pose estimation. • Hybrid-kinematical feature encoder extracts kinematical features of 2D pose. • Spatial–temporal feature transform enhances the spatial and temporal correlations. • The fusion of spatial and temporal features promote the final 3D pose estimation. [ABSTRACT FROM AUTHOR]

Video-based 3D human pose estimation is an important yet challenging task for many human-involved pattern recognition systems. Existing deep learning-based 3D human pose estimation methods are faced with the problems of lacking large-scale training data and lacking effective solutions to represent the complicated human body structure. To this end, this paper proposes a jo intl y learning framework entitled JoyPose that simultaneously leverages both human pose data augmentation and human pose estimation. In particular, JoyPose consists of an evolutionary data augmentation module and an anatomy-aware global–local pose feature representation module for 3D human pose estimation. The evolution for data augmentation is guided by a reinforcement learning strategy in a probabilistic way according to pose estimation loss. The anatomy-aware global–local pose feature representation module separately captures global features and local features according to anatomical and kinematic patterns observed from pose estimation errors across different human joints. The performance of the final human pose estimation is leveraged by both data augmentation and anatomy-aware global–local feature representation. Extensive experiments on three real-world datasets demonstrate the superiority and robustness against state-of-the-art methods. • JoyPose simultaneously leverages 3D human pose augmentation and pose estimation. • The distributions of crossover and mutation are learned for human pose augmentation. • Pose estimation quality is utilized to guide the distribution updating. • Global and local features are extracted based on the kinematical characters. [ABSTRACT FROM AUTHOR]

Distributed scheduling with assembly machines has been an attractive research field in sustainable supply chains and multi-factory manufacturing systems. This paper investigates a distributed blocking flow shop scheduling problem with an assembly machine (DABFSP) with the total assembly completion time criterion, and proposes an effective discrete monarch butterfly optimization algorithm (EDMBO). First, a constructive heuristic combining the largest processing time rule and the earliest start assembly rule is provided to find a promising sequence. On this basis, an efficient initialization method is introduced to generate an initial population with high quality and diversity. Afterward, a global search procedure is presented, which integrates four kinds of improved operators and expands the solution space in a good direction. Then, according to different problem-specific characteristics, we present four targeted and flexible variable neighborhood search methods based on the critical job and critical factory to exploit the solution space. Finally, statistically significant numerical experiments are carried out with state-of-the-art optimization methods based on 1710 benchmark instances. The experimental results and detailed analysis demonstrate that the EDMBO is superior to preferred algorithms for addressing the DABFSP. [ABSTRACT FROM AUTHOR]

Automatic classification of texture images is an important and challenging task in the applications of image analysis and scene understanding. In this paper, we focus on the problem of the classification of texture images acquired under various rotation and illumination conditions and propose a new local image descriptor which is named local spiking pattern (LSP). Specifically, the proposed LSP uses a 2-dimensional neural network, which is made up of a series of interconnected spiking neurons, to generate binary images by iteration. The binary images are then encoded to generate discriminative feature vectors. In classification phase, we use a nearest neighborhood classifier to achieve supervised classification. Finally, LSP is evaluated by comparison with some state-of-the-art local image descriptors. Experimental results on Outex texture database show that LSP outperforms most of the other local image descriptors in the noiseless case and shows high robustness when texture images are distorted by salt & pepper noise. [ABSTRACT FROM AUTHOR]

As a matter of fact, mammalian visual system do not pay an equivalent attention to different regions in an image, the visual cortex is less sensitive to textures than non-textures. Therefore, to obtain the optimal visual quality and the perfect compression ratio simultaneously in image quantization, textures should be quantized coarsely, and non-textures should be quantized finely. The pulse-coupled neural networks (PCNN) is a model of synchronous pulse bursts in mammalian visual cortex, which has been proved to be extremely effective in image processing because of its biological background. In this work, a mammalian visual characteristics inspired perceptual image quantization strategy is proposed. It employs PCNN to extract textures from original image. Then, pixels in textures are quantized into less gray scale layers than pixels in non-textures. After that, quantized textures and quantized non-textures are consolidated. Experimental results prove validity and efficiency of the proposed method. [ABSTRACT FROM AUTHOR]

Abstract: Background: As drug-eluting stent (DES) has almost overcome the disadvantage of frequent restenosis, off-pump coronary artery bypass grafting (OPCAB) has been introduced to avoid complications of cardiopulmonary bypass. However, which approach may promise better outcomes for patients with coronary artery disease remains controversial. Methods: Three databases were searched. The outcomes of interest were major adverse cardiac and cerebrovascular events (MACCE), all-cause death, target vessel revascularization (TVR), repeat revascularization (RRV), myocardial infarction (MI), and cerebrovascular events (CVE). The relative risk (RR) was calculated as the summary statistic. Results: 11,452 patients from 22 studies were included, of which 4949 patients underwent OPCAB and 6503 patients received DES. The cumulative rates of MACCE (RR [95% CI]=0.43 [0.34, 0.54], P <0.00001), all-cause death (RR [95% CI]=0.56 [0.33, 0.96], P =0.03), TVR (RR [95% CI]=0.33 [0.21, 0.53], P <0.00001), RRV (RR [95% CI]=0.22 [0.11, 0.42], P <0.00001) and MI (RR [95% CI]=0.13 [0.05, 0.29], P <0.00001) at 3years were all lower in OPCAB group. The incidences of in-hospital death (RR [95% CI]=1.31 [0.81, 2.13], P =0.27) and MI (RR [95% CI]=1.03 [0.60, 1.78], P =0.92) were not different between groups, but the rate of in-hospital CVE was lower (RR [95% CI]=2.6355 [1.0033, 6.9228], P =0.05) in DES group. Conclusions: OPCAB presents better long-term outcomes of MACCE, all-cause mortality, TVR, RRV and MI but uncertain outcome of postoperative CVE without influencing the incidences of in-hospital death and MI. [Copyright &y& Elsevier]