Scene text detection and recognition have been well explored in the past few years. Despite the progress, efficient and accurate end-to-end spotting of arbitrarily-shaped text remains challenging. In ...this work, we propose an end-to-end text spotting framework, termed PAN++, which can efficiently detect and recognize text of arbitrary shapes in natural scenes. PAN++ is based on the kernel representation that reformulates a text line as a text kernel (central region) surrounded by peripheral pixels. By systematically comparing with existing scene text representations, we show that our kernel representation can not only describe arbitrarily-shaped text but also well distinguish adjacent text. Moreover, as a pixel-based representation, the kernel representation can be predicted by a single fully convolutional network, which is very friendly to real-time applications. Taking the advantages of the kernel representation, we design a series of components as follows: 1) a computationally efficient feature enhancement network composed of stacked Feature Pyramid Enhancement Modules (FPEMs); 2) a lightweight detection head cooperating with Pixel Aggregation (PA); and 3) an efficient attention-based recognition head with Masked RoI. Benefiting from the kernel representation and the tailored components, our method achieves high inference speed while maintaining competitive accuracy. Extensive experiments show the superiority of our method. For example, the proposed PAN++ achieves an end-to-end text spotting F-measure of 64.9 at 29.2 FPS on the Total-Text dataset, which significantly outperforms the previous best method. Code will be available at: git.io/PAN .
Scene text detection is an important step in the scene text reading system. The main challenges lie in significantly varied sizes and aspect ratios, arbitrary orientations, and shapes. Driven by the ...recent progress in deep learning, impressive performances have been achieved for multi-oriented text detection. Yet, the performance drops dramatically in detecting the curved texts due to the limited text representation (e.g., horizontal bounding boxes, rotated rectangles, or quadrilaterals). It is of great interest to detect the curved texts, which are actually very common in natural scenes. In this paper, we present a novel text detector named TextField for detecting irregular scene texts. Specifically, we learn a direction field pointing away from the nearest text boundary to each text point. This direction field is represented by an image of 2D vectors and learned via a fully convolutional neural network. It encodes both binary text mask and direction information used to separate adjacent text instances, which is challenging for the classical segmentation-based approaches. Based on the learned direction field, we apply a simple yet effective morphological-based post-processing to achieve the final detection. The experimental results show that the proposed TextField outperforms the state-of-the-art methods by a large margin (28% and 8%) on two curved text datasets: Total-Text and SCUT-CTW1500, respectively; TextField also achieves very competitive performance on multi-oriented datasets: ICDAR 2015 and MSRA-TD500. Furthermore, TextField is robust in generalizing unseen datasets.
•A novel strategy for damage detection and localization based on Lamb wave focusing and dispersion compensation.•Numerical and experimental verification of the proposed method.•Comparison of the ...proposed method with delay-and-sum algorithm.
Piezoelectric transducer arrays are utilized in Structural Health Monitoring systems as a means for excitation and sensing of elastic waves. Anomalies of propagating waves have enabled to develop damage detection algorithms. Depending on actuation-sensing strategies these algorithms can be classified as pitch-catch and pulse-echo. Despite many signal processing methods such as delay-sum, time-reversal, probability-based diagnostic imaging, etc. the spatial damage information provided by the actuator-sensor paths to reconstruct the damage image is limited. A novel strategy based on Lamb wave focusing is proposed in order to increase damage imaging resolution. In the proposed method all actuators are used at the same time exciting specially designed waveforms so that inspect one specific point of the structure. Damage map is created by applying appropriate signal processing. It uses dispersion curve of selected Lamb wave mode for dispersion compensation. The dispersion curve is acquired by using laser scanning Doppler vibrometer. The damage indicator is calculated based on the energy of compensated signals registered by sensors. It is shown that apart from high energy level at excitation point, energy is concentrated exactly in the damaged region. An example of crack detection and visualization in an aluminum plate is shown confirming the accuracy of the proposed method. Also the proposed method is compared to well-established delay-and-sum algorithm.
The ultrasonic guided wave has emerged as one of the most prominent and promising tools for metal and composite structures in the fields of structural health monitoring (SHM) and nondestructive ...testing (NDT). This paper presents a novel model-based 2D multiple signal classification (MUSIC) damage identification algorithm for plate-like structures. Unlike the conventional MUSIC algorithm, the proposed model-based 2D MUSIC damage identification algorithm is deduced based on the assumption of near-field according to the propagation model of guided waves. Since scattered signals contain the location information of damage, the cross-correlation function of residual signals received by experiment and scattered signals received by damage scattering model are developed for spatial spectrum estimation MUSIC algorithm. Due to the uncorrelation of signal and noise, the damage can be successfully identified by searching the peak point of spatial spectrum in the monitored area employing the orthogonality of signal subspace and noise subspace. The accuracy and effectiveness of the proposed method are firstly validated by numerical simulation on aluminum plate, and the general applicability is further verified by experiments for the damage identification of laminated composite plate. The numerical and experimental results demonstrate the proposed damage identification algorithm is appropriate for damage identification of plate-like structures with high estimation accuracy and resolution.
The digital economy and ecological environment are two major issues related to high-quality economic development. Scholars have not yet reached a unified conclusion about the link between the digital ...economy and pollution emissions, and the impact mechanism of the former on the latter needs further study. Using data from 278 Chinese cities from 2010 to 2019, this research employs coupling coordination analysis, fixed effect analysis and mediation analysis to examine the heterogeneous impact mechanisms of the expansion of the digital economy on urban pollution reduction from many angles. It discovers that, first, the growth of the digital economy has decreased the discharge of urban pollutants overall. Second, the impact mechanisms of the digital economy are heterogeneous. From a regional perspective, industrial structure supererogation plays an intermediary role in the relationship between digital economy development and pollution reduction in the eastern and central regions, but the mediating effect is not significant in the western and northeastern regions. In terms of the city development level, industrial structure supererogation has significantly mediated the relationship between the growth of the digital economy and the reduction of pollution in first- and second-tier cities, but this mediating effect is not significant in third-tier and other cities. Third, the above conclusions are still valid after the robustness test is carried out using instrumental variable estimation, replacement of the estimation method, and replacement of explanatory variables. This study is a useful contribution to research on the effects of the digital economy and the factors influencing pollution reduction. The results advance the study of the digital economy and also have practical implications for improving China's ecological environment and fostering high-quality economic growth. Finally, we provide policy suggestions for the coordinated promotion of the digital economy's development, industrial structure supererogation and environmental pollution reduction.
Abstract
Artificial intelligence (AI) has been widely applied in drug discovery with a major task as molecular property prediction. Despite booming techniques in molecular representation learning, ...key elements underlying molecular property prediction remain largely unexplored, which impedes further advancements in this field. Herein, we conduct an extensive evaluation of representative models using various representations on the MoleculeNet datasets, a suite of opioids-related datasets and two additional activity datasets from the literature. To investigate the predictive power in low-data and high-data space, a series of descriptors datasets of varying sizes are also assembled to evaluate the models. In total, we have trained 62,820 models, including 50,220 models on fixed representations, 4200 models on SMILES sequences and 8400 models on molecular graphs. Based on extensive experimentation and rigorous comparison, we show that representation learning models exhibit limited performance in molecular property prediction in most datasets. Besides, multiple key elements underlying molecular property prediction can affect the evaluation results. Furthermore, we show that activity cliffs can significantly impact model prediction. Finally, we explore into potential causes why representation learning models can fail and show that dataset size is essential for representation learning models to excel.
We have developed a new mass spectrometry (MS) technology, the Single-probe MS, capable of real-time, in situ metabolomic analysis of individual living cells. The Single-probe is a miniaturized ...multifunctional sampling and ionization device that is directly coupled to the mass spectrometer. With a sampling tip smaller than individual eukaryotic cells (<10 μm), the Single-probe can be inserted into single cells to sample the intracellular compounds for real-time MS analysis. We have used the Single-probe to detect several cellular metabolites and the anticancer small molecules paclitaxel, doxorubicin, and OSW-1 in individual cervical cancer cells (HeLa). Single cell mass spectrometry (SCMS) is an emerging scientific technology that could reshape the analytical science of many research disciplines, and the Single-probe MS technology is a novel method for SCMS that, through its accessible fabrication protocols, can be broadly applied to different research areas.
Traditional approaches for the assessment of physiological responses of microbes in the environment rely on bulk filtration techniques that obscure differences among populations as well as among ...individual cells. Here, were report on the development on a novel micro-scale sampling device, referred to as the "Single-probe," which allows direct extraction of metabolites from living, individual phytoplankton cells for mass spectrometry (MS) analysis. The Single-probe is composed of dual-bore quartz tubing which is pulled using a laser pipette puller and fused to a silica capillary and a nano-ESI. For this study, we applied Single-probe MS technology to the marine dinoflagellate
, assaying cells grown under different illumination levels and under nitrogen (N) limiting conditions as a proof of concept for the technology. In both experiments, significant differences in the cellular metabolome of individual cells could readily be identified, though the vast majority of detected metabolites could not be assigned to KEGG pathways. Using the same approach, significant changes in cellular lipid complements were observed, with individual lipids being both up- and down-regulated under light vs. dark conditions. Conversely, lipid content increased across the board under N limitation, consistent with an adjustment of Redfield stoichiometry to reflect higher C:N and C:P ratios. Overall, these data suggest that the Single-probe MS technique has the potential to allow for near
metabolomic analysis of individual phytoplankton cells, opening the door to targeted analyses that minimize cell manipulation and sampling artifacts, while preserving metabolic variability at the cellular level.
Despite the development of various methods and commercial kits, site-directed mutagenesis of large plasmids remains a challenge in many laboratories. A site-directed mutagenesis method was ...developed for large plasmids by directly transforming two overlapping PCR fragments into
. This method successfully generated mutations for plasmids of 8.3 kb and 11.0 kb with high efficiencies. The method only requires Q5 DNA polymerase and
, which greatly reduces costs. The procedure is simple, including PCR reaction,
treatment and transformation. This simple, efficient and economical site-directed mutagenesis method for large plasmids is likely to be widely applied in the future.
A simple, efficient and economical site-directed mutagenesis method was developed for large plasmids by directly transforming two overlapping PCR fragments. The method only requires Q5 DNA polymerase and
. Researchers successfully generated mutations for plasmids up to 11 kb.