Institution
Naver Corporation
Company•Seongnam-si, South Korea•
About: Naver Corporation is a company organization based out in Seongnam-si, South Korea. It is known for research contribution in the topics: Terminal (electronics) & Computer science. The organization has 4038 authors who have published 4294 publications receiving 35045 citations. The organization is also known as: NAVER Corporation & NAVER.
Papers published on a yearly basis
Papers
More filters
•
TL;DR: A new scene text detection method to effectively detect text area by exploring each character and affinity between characters by exploiting both the given character- level annotations for synthetic images and the estimated character-level ground-truths for real images acquired by the learned interim model.
Abstract: Scene text detection methods based on neural networks have emerged recently and have shown promising results. Previous methods trained with rigid word-level bounding boxes exhibit limitations in representing the text region in an arbitrary shape. In this paper, we propose a new scene text detection method to effectively detect text area by exploring each character and affinity between characters. To overcome the lack of individual character level annotations, our proposed framework exploits both the given character-level annotations for synthetic images and the estimated character-level ground-truths for real images acquired by the learned interim model. In order to estimate affinity between characters, the network is trained with the newly proposed representation for affinity. Extensive experiments on six benchmarks, including the TotalText and CTW-1500 datasets which contain highly curved texts in natural images, demonstrate that our character-level text detection significantly outperforms the state-of-the-art detectors. According to the results, our proposed method guarantees high flexibility in detecting complicated scene text images, such as arbitrarily-oriented, curved, or deformed texts.
35 citations
••
TL;DR: In this article, the authors presented the modeling of an augmented HVAC system including CO2 concentration, and its control strategies, and a linear-quadratic regulator (LQR) was designed to optimize control performance and stabilize the proposed system.
Abstract: Conventional heating, ventilating, and air conditioning (HVAC) systems have traditionally used the temperature and the humidity ratio as the quantitative indices of comfort in a room. Recently, the carbon dioxide (CO2) concentration has also been recognized as having an important contribution to room comfort. This paper presents the modeling of an augmented HVAC system including CO2 concentration, and its control strategies. Because the proposed augmented HVAC system is multi-input multi-output (MIMO) and has no relative degree problem, the dynamic extension algorithm can be employed; then, a feedback linearization technique is applied. A linear-quadratic regulator (LQR) is designed to optimize control performance and to stabilize the proposed HVAC system. Simulation results are provided to validate the proposed system model, as well as its linearized control system.
35 citations
••
01 Sep 2019TL;DR: Experimental results show that the proposed ExcitNet vocoder, trained both speaker-dependently and speaker-independently, outperforms traditional linear prediction vocoders and similarly configured conventional WaveNet Vocoders.
Abstract: This paper proposes a WaveNet-based neural excitation model (ExcitNet) for statistical parametric speech synthesis systems. Conventional WaveNet-based neural vocoding systems significantly improve the perceptual quality of synthesized speech by statistically generating a time sequence of speech waveforms through an auto-regressive framework. However, they often suffer from noisy outputs because of the difficulties in capturing the complicated time-varying nature of speech signals. To improve modeling efficiency, the proposed ExcitNet vocoder employs an adaptive inverse filter to decouple spectral components from the speech signal. The residual component, i.e. excitation signal, is then trained and generated within the WaveNet framework. In this way, the quality of the synthesized speech signal can be further improved since the spectral component is well represented by a deep learning framework and, moreover, the residual component is efficiently generated by the WaveNet framework. Experimental results show that the proposed ExcitNet vocoder, trained both speaker-dependently and speaker-independently, outperforms traditional linear prediction vocoders and similarly configured conventional WaveNet vocoders.
34 citations
••
TL;DR: In this article, a synthetic scheme was developed for the large-scale preparation of a dimethylthiophene-fused and tetrahydroquinaldine-linked dimethylcyclopentadienyl titanium complex (2), which is a highperformance homogeneous Ziegler catalyst.
Abstract: A synthetic scheme was developed for the large-scale preparation of a dimethylthiophene-fused and tetrahydroquinaldine-linked dimethylcyclopentadienyl titanium complex (2), which is a high-performance homogeneous Ziegler catalyst. 2,3,4,5-Tetramethyl-4,5-dihydrocyclopenta[b]thiophen-6-one was prepared without chromatography purification on the 40-g scale in a laboratory setting, from which the ligand precursor for 2 was obtained in 65% yield on a 50-g scale in a one-pot without the need for chromatography purification. Metallation was achieved in a high yield (78%) through reaction of the dilithiated compound with TiCl4. Many derivatives were prepared by employing the same synthetic scheme as applied for 2. Among them, the titanium complex prepared from 2-methyl-4,5-dimethyl-6-(2-n-butyl-2,3,4,5-tetrahydroquinolin-8-yl)-4H-cyclopenta[b]thiophene exhibited an exceptionally high activity. Under commercially relevant high-temperature polymerization conditions (160 °C), this compound showed a higher activity than 2 (126 × 106 g/molTi∙h versus 72 × 106 g/molTi∙h), albeit with the formation of a polymer of slightly lower molecular weight (Mw, 159,000 versus 218,000) and with a slightly lower 1-octene content (9.3 mol% versus 12 mol%).
34 citations
•
12 Jul 2020TL;DR: In this paper, the authors proposed density and coverage metrics for image generation, which provide more interpretable and reliable signals for practitioners than the existing metrics, such as precision and recall.
Abstract: Devising indicative evaluation metrics for the image generation task remains an open problem. The most widely used metric for measuring the similarity between real and generated images has been the Frechet Inception Distance (FID) score. Because it does not differentiate the fidelity and diversity aspects of the generated images, recent papers have introduced variants of precision and recall metrics to diagnose those properties separately. In this paper, we show that even the latest version of the precision and recall metrics are not reliable yet. For example, they fail to detect the match between two identical distributions, they are not robust against outliers, and the evaluation hyperparameters are selected arbitrarily. We propose density and coverage metrics that solve the above issues. We analytically and experimentally show that density and coverage provide more interpretable and reliable signals for practitioners than the existing metrics. Code: this https URL.
34 citations
Authors
Showing all 4041 results
Name | H-index | Papers | Citations |
---|---|---|---|
Andrea Vedaldi | 89 | 305 | 63305 |
Sunghun Kim | 51 | 115 | 12994 |
Eric Gaussier | 41 | 231 | 8203 |
Un Ju Jung | 39 | 98 | 5696 |
Hyun-Soo Kim | 37 | 421 | 5650 |
Gabriela Csurka | 37 | 145 | 10959 |
Nojun Kwak | 34 | 234 | 6026 |
Young-Jin Park | 31 | 257 | 3759 |
Sung Joo Kim | 31 | 196 | 3078 |
Jae-Hoon Kim | 30 | 323 | 5847 |
Jung-Ryul Lee | 29 | 222 | 3322 |
Joon Son Chung | 28 | 73 | 4900 |
Ok-Hwan Lee | 27 | 163 | 2896 |
Diane Larlus | 27 | 69 | 4722 |
Jung Goo Lee | 26 | 142 | 1917 |