Connectionist Temporal Classification (CTC) and attention mechanism are two main approaches used in recent scene text recognition works. Compared with attention-based methods, CTC decoder has a much shorter inference time, yet a lower accuracy. To design an efficient and effective model, we propose the guided training of CTC (GTC), where CTC model learns a better alignment and feature representations from a more powerful attentional guidance. With the benefit of guided training, CTC model achieves robust and accurate prediction for both regular and irregular scene text while maintaining a fast inference speed. Moreover, to further leverage the potential of CTC decoder, a graph convolutional network (GCN) is proposed to learn the local correlations of extracted features. Extensive experiments on standard benchmarks demonstrate that our end-to-end model achieves a new state-of-the-art for regular and irregular scene text recognition and needs 6 times shorter inference time than attention-based methods.

No abstract

The three-dimensional finite difference method was used in this study to analyze the deformation and stresses of a passive pile under surcharge load in extensively deep soft soil. A three-dimensional numerical model was proposed and verified by a field test. The horizontal displacements of the pile agreed well with the field results. This study investigated the pile-foundation soil interaction, the load transfer mechanism, the excess pore water pressure (EPWP), and the horizontal resistance of the foundation soil. The results show that the soil in the corner of the loading area developed a large uplift deformation, while the center of the loading area developed a large settlement. The lateral displacement of the pile decreased sharply with the increase of the depth and increased with the surcharge load. The lateral displacement of the soil was negligible when the depth exceeded 30 m. The EPWP increased in a nonlinear way with the increase of the surcharge load and accumulated with the placement of the new lift. The distribution of the lateral earth pressure in the shallow soil layer was complex, and the negative value was observed under a high surcharge load due to the suction effect. The proportion coefficient of the horizontal resistance coefficient showed much smaller value in the situation of large lateral deformation and high surcharge load. The design code overestimated the horizontal resistance of the shallow foundation soil, which should be given attention for the design and analysis of the laterally loaded structures in extensively soft soil.

Background: With advances in medicine, there have been more and more ways to treat renal calculi in recent years. Percutaneous nephrolithotomy (PCNL) is a safe and effective treatment. The purpose of this paper is to study the efficacy and safety of PCNL combined with negative pressure suction in the treatment of renal calculi by meta-analysis.

Connectionist Temporal Classification (CTC) and attention mechanism are two main approaches used in recent scene text recognition works. Compared with attention-based methods, CTC decoder has a much shorter inference time, yet a lower accuracy. To design an efficient and effective model, we propose the guided training of CTC (GTC), where CTC model learns a better alignment and feature representations from a more powerful attentional guidance. With the benefit of guided training, CTC model achieves robust and accurate prediction for both regular and irregular scene text while maintaining a fast inference speed. Moreover, to further leverage the potential of CTC decoder, a graph convolutional network (GCN) is proposed to learn the local correlations of extracted features. Extensive experiments on standard benchmarks demonstrate that our end-to-end model achieves a new state-of-the-art for regular and irregular scene text recognition and needs 6 times shorter inference time than attentionbased methods. * This work was done when Wenyang was an intern at Sense-Time Group Ltd.