Scene text recognition is a hot research topic in computer vision. Recently, many methods based on the encoder-decoder framework have been proposed, and they can handle scene texts of perspective distortion curve shape. Nevertheless, still face lots challenges like image blur, uneven illumination, incomplete characters. We argue that most are local visual features without explicit global semantic information. In this work, we propose semantics enhanced to robustly recognize low-quality...

In self-supervised spatio-temporal representation learning, the temporal resolution and long-short term characteristics are not yet fully explored, which limits capabilities of learned models. this paper, we propose a novel method, referred to as video Playback Rate Perception (PRP), learn in simple-yet-effective way. PRP roots dilated sampling strategy, produces self-supervision signals about playback rates for model learning. is implemented with feature encoder, classification module,...

We propose a novel self-supervised method, referred to as Video Cloze Procedure (VCP), learn rich spatial-temporal representations. VCP first generates “blanks” by withholding video clips and then creates “options” applying spatio-temporal operations on the withheld clips. Finally, it fills blanks with learns representations predicting categories of applied can act either proxy task or target in learning. As task, converts into clip (options), which enhances flexibility reduces complexity...

Large language models (LLMs), such as ChatGPT, are able to generate human-like, fluent responses for many downstream tasks, e.g., task-oriented dialog and question answering. However, applying LLMs real-world, mission-critical applications remains challenging mainly due their tendency hallucinations inability use external knowledge. This paper proposes a LLM-Augmenter system, which augments black-box LLM with set of plug-and-play modules. Our system makes the grounded in knowledge, stored...

In this paper, we present our work in emotion cause extraction.Since there is no open dataset available, the lack of annotated resources has limited research area.Thus, first a built using SINA city news.The annotation based on scheme W3C Emotion Markup Language.Second, propose 7-tuple definition to describe events.Based general definition, new event-driven extraction method multi-kernel SVMs where syntactical tree approach used represent events text.A convolution kernel multikernel SVM are...

This work investigates the aerodynamics of a NACA 0012 airfoil at chord-based Reynolds numbers (Rec) from 5.3 × 103 to 2.0 104. The lift and drag coefficients, CL CD, airfoil, along with flow structure, were measured as turbulent intensity Tu oncoming varies 0.6% 6.0%. analysis present data those in literature unveils total eight distinct structures around suction side airfoil. Four Rec regimes, i.e., ultra-low (<1.0 104), low (1.0 104–3.0 105), moderate (3.0 105–5.0 106), high...

Junnan Zhu, Qian Wang, Yining Yu Zhou, Jiajun Zhang, Shaonan Chengqing Zong. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.

This paper presents a systematic study of the cross-flow-induced vibration on spring-supported circular cylinder diameter $D$ placed in wake fixed smaller $d$ . The ratios $d/D$ and $L/d$ are varied from 0.2 to 1.0 5.5, respectively, where $L$ is distance between centre upstream forward stagnation point downstream cylinder. Extensive measurements conducted capture frequency responses, surface pressure, shedding frequencies flow fields using laser vibrometer, hot-wire, pressure scanner...

This work aims to provide a systematic experimental study on the wake of two tandem cylinders unequal diameters. The fluid dynamics around circular cylinder diameter $D$ placed in another with smaller $d$ is investigated, including time-mean drag coefficient ( $C_{D}$ ), fluctuating and lift coefficients $C_{D}^{\prime }$ $C_{L}^{\prime Strouhal number $St$ ) flow structures. Reynolds based kept constant at $4.27\times 10^{4}$ . ratios $d/D$ $L/d$ vary from 0.2 1.0 8.0 respectively, where...

Multimodal summarization with multimodal output (MSMO) is to generate a summary for news report, which has been proven effectively improve users' satisfaction. The existing MSMO methods are trained by the target of text modality, leading modality-bias problem that ignores quality model-selected image during training. To alleviate this problem, we propose objective function guidance reference use loss from generation and selection. Due lack data, present two strategies, i.e., ROUGE-ranking...

Nowadays, scene text recognition has attracted more and attention due to its various applications. Most state-of-the-art methods adopt an encoder-decoder framework with mechanism, which generates autoregressively from left right. Despite the convincing performance, speed is limited because of one-by-one decoding strategy. As opposed autoregressive models, non-autoregressive models predict results in parallel a much shorter inference time, but accuracy falls behind counterpart considerably....

The combination of optogenetics and electrophysiological recording enables high-precision bidirectional interactions between neural interfaces circuits, which provides a promising approach for the study progressive neurophysiological phenomena. Opto-electrophysiological probes with sufficient flexibility biocompatibility are desirable to match low mechanical stiffness brain tissue chronic reliable performance. However, lack rigidity poses challenges accurate implantation flexible less...

With the continuous advancement of socio-economic levels and relentless innovation in modern medical technologies, there's been a significant increase importance people place on their physiological health, particularly context colorectal cancer—a prevalent malignant tumor that has captivated widespread attention within community for its prevention treatment. Notably, polyps, identified as precursors to cancer, are crucial early diagnosis precise detection, serving fundamental elements...

Traffic classification, a mapping of traffic to network applications, is important for variety networking and security issues, such as measurement, monitoring, well the detection malware activities. In this paper, we propose Securitas, trace-based protocol identification system, which exploits semantic information in message formats. Securitas requires no prior knowledge specifications. Deeming language between two processes, our approach based upon new insight that n-grams traces, just like...

Abstract Active control of a turbulent boundary layer has been experimentally investigated with view to reducing the skin-friction drag and gaining some insight into mechanism that leads reduction. A spanwise-aligned array piezo-ceramic actuators was employed generate transverse travelling wave along wall surface, specified phase shift between adjacent actuators. Local exhibits strong dependence on parameters, including wavelength, amplitude frequency oscillation. maximum reduction 50 %...

Cross-language multidocument summarization is the task to generate a summary in target language (e.g., Chinese) from collection of documents different source English). Previous methods such as extractive and compressive algorithms focus only on single sentence selection compression, which cannot make full use similar sentences containing complementary information. Furthermore, translation model knowledge not fully explored previous approaches. To address these two problems, we propose this...