This paper presents a novel convolutional neural network (CNN) -based method for high-accuracy real-time car license plate detection. Many contemporary methods detection are reasonably effective under the specific conditions or strong assumptions only. However, they exhibit poor performance when assessed images have degree of rotation, as result manual capture by traffic police deviation camera. Therefore, we propose CNN-based MD-YOLO framework multi-directional Using accurate rotation angle...

In this paper, we propose a novel method, aggregation cross-entropy (ACE), for sequence recognition from brand new perspective. The ACE loss function exhibits competitive performance to CTC and the attention mechanism, with much quicker implementation (as it involves only four fundamental formulas), faster inference\back-propagation (approximately O(1) in parallel), less storage requirement (no parameter negligible runtime memory), convenient employment (by replacing ACE). Furthermore,...

Scene text in the wild is commonly presented with high variant characteristics. Using quadrilateral bounding box to localize instance nearly indispensable for detection methods. However, recent researches reveal that introducing scene will bring a label confusion issue which easily overlooked, and this may significantly undermine performance. To address issue, paper, we propose novel method called Sequential-free Box Discretization (SBD) by discretizing into key edges (KE) can further derive...

This paper presents a method that can accurately detect heads especially small under the indoor scene. To achieve this, we propose novel method, Feature Refine Net (FRN), and cascaded multi-scale architecture. FRN exploits hierarchical features created by deep convolutional neural networks. The proposed channel weighting enables to make use of alternatively effectively. improve performance head detection, architecture which has two detectors. One called global detector is responsible for...

Most current detection methods have adopted anchor boxes as regression references. However, the performance is sensitive to setting of boxes. A proper may vary significantly across different datasets, which severely limits universality detectors. To improve adaptivity detectors, in this paper, we present a novel dimension-decomposition region proposal network (DeRPN) that can perfectly displace traditional Region Proposal Network (RPN). DeRPN utilizes an string mechanism independently match...

Visual Information Extraction (VIE) task aims to extract key information from multifarious document images (e.g., invoices and purchase receipts). Most previous methods treat the VIE simply as a sequence labeling problem or classification problem, which requires models carefully identify each kind of semantics by introducing multimodal features, such font, color, layout. But features can't work well when faced with numeric semantic categories some ambiguous texts. To address this issue, in...

Evaluation protocols play key role in the developmental progress of text detection methods. There are strict requirements to ensure that evaluation methods fair, objective and reasonable. However, existing metrics exhibit some obvious drawbacks: 1) They not goal-oriented; 2) they cannot recognize tightness methods; 3) one-to-many many-to-one solutions involve inherent loopholes deficiencies. Therefore, this paper proposes a novel protocol called Tightness-aware Intersect-over-Union (TIoU)...

The alpine sandy dune ecosystem is highly vulnerable to global climate change. Ecological stoichiometry in plants and soils plays a crucial role biogeochemical cycles, energy flow functioning ecosystems. However, the stoichiometric changes correlations of among different types dunes have not been fully explored. Three (moving dune, MD; semifixed SFD; fixed FD) Sophora moorcroftiana shrub middle reaches Yarlung Zangbo River were used as subjects current study. Plant community characteristics,...

Abstract Modeling in computer vision has long been dominated by convolutional neural networks (CNNs). Recently, light of the excellent performance self-attention mechanism language field, transformers tailored for visual data have drawn significant attention and triumphed over CNNs various tasks. These heavily rely on large-scale pre-training to achieve competitive accuracy, which not only hinders freedom architectural design downstream tasks like object detection, but also causes learning...

Abstract The primary objective of multi-objective evolutionary algorithms (MOEAs) is to find a set evenly distributed nondominated solutions that approximate the Pareto front (PF) optimization problem (MOP) or many-objective (MaOP). This implies approximated solution obtained by MOEAs should be as close PF possible while remaining diverse, adhering criteria convergence and diversity. However, existing exhibit an imbalance between achieving maintaining diversity in space. As far criterion...

Abstract We added jasplakinolide to anaphase crane‐fly spermatocytes and determined its effects on chromosome movement. Previous work showed that the actin depolymerizing agents cytochalasin D or latrunculin B blocked slowed movements. studied of jasplakinolide, a compound stabilizes filaments. Jasplakinolide had same effect movements each half‐ bivalent in separating pair half‐bivalents, but different half‐bivalent pairs cell often responded differently, even when concentrations varied by...

Scene text in the wild is commonly presented with high variant characteristics. Using quadrilateral bounding box to localize instance nearly indispensable for detection methods. However, recent researches reveal that introducing scene will bring a label confusion issue which easily overlooked, and this may significantly undermine performance. To address issue, paper, we propose novel method called Sequential-free Box Discretization (SBD) by discretizing into key edges (KE) can further derive...

Abstract A method for the bromination of α ‐diazo phenylacetate derivatives using cobalt(II) bromide is described. This reaction features a short time, broad substrate scope, operational simplicity, acid‐free conditions, and gram‐scalability. magnified image

Pseudo-Labeling based semi-supervised learning has shown promising advantages in Scene Text Recognition (STR). Most of them usually use a pre-trained model to generate sequence-level pseudo labels for text images and then re-train the model. Recently, conducting teacher-student framework (a student is supervised by from teacher model) become increasingly popular, which trains an end-to-end manner yields outstanding performance learning. However, applying this directly STR exhibits unstable...

Overgrazing leads to grassland degradation and productivity decline. Rest-grazing during the regreen-up period can quickly restore fertilization is a common restoration strategy. However, effects of rest-grazing time on soil microorganisms are unclear in alpine grasslands. Therefore, experiment was carried out explore response measures. A field control with as factors have been conducted from when grass returned green till livestock moved summer pasture Dawu Town Maqin County China. The...

This paper presents a method that can accurately detect heads especially small under the indoor scene. To achieve this, we propose novel method, Feature Refine Net (FRN), and cascaded multi-scale architecture. FRN exploits hierarchical features created by deep convolutional neural networks. The proposed channel weighting enables to make use of alternatively effectively. improve performance head detection, architecture which has two detectors. One called global detector is responsible for...

Most current detection methods have adopted anchor boxes as regression references. However, the performance is sensitive to setting of boxes. A proper may vary significantly across different datasets, which severely limits universality detectors. To improve adaptivity detectors, in this paper, we present a novel dimension-decomposition region proposal network (DeRPN) that can perfectly displace traditional Region Proposal Network (RPN). DeRPN utilizes an string mechanism independently match...

Evaluation protocols play key role in the developmental progress of text detection methods. There are strict requirements to ensure that evaluation methods fair, objective and reasonable. However, existing metrics exhibit some obvious drawbacks: 1) They not goal-oriented; 2) they cannot recognize tightness methods; 3) one-to-many many-to-one solutions involve inherent loopholes deficiencies. Therefore, this paper proposes a novel protocol called Tightness-aware Intersect-over-Union (TIoU)...

This study constructs a cloud computing-based college English multimedia test question modeling and application through an in-depth of computing questions. The emergence technology undoubtedly provides new ideal method to solve data paper management problems. analyzes the advantages Hadoop platform MapReduce model builds distributed based on using universities’ existing hardware software resources. model. UML system is given, implemented, tested functionally, results analysis are given....

Integrating first-order logic constraints (FOLCs) with neural networks is a crucial but challenging problem since it involves modeling intricate correlations to satisfy the constraints. This paper proposes novel layer, LogicMP, whose layers perform mean-field variational inference over an MLN. It can be plugged into any off-the-shelf network encode FOLCs while retaining modularity and efficiency. By exploiting structure symmetries in MLNs, we theoretically demonstrate that our well-designed,...