• Title/Summary/Keyword: Weak Label

Search Result 12, Processing Time 0.022 seconds

An Analysis of the methods to alleviate the cost of data labeling in Deep learning (딥 러닝에서 Labeling 부담을 줄이기 위한 연구분석)

  • Han, Seokmin
    • The Journal of the Convergence on Culture Technology
    • /
    • v.8 no.1
    • /
    • pp.545-550
    • /
    • 2022
  • In Deep Learning method, it is well known that it requires large amount of data to train the deep neural network. And it also requires the labeling of each data to fully train the neural network, which means that experts should spend lots of time to provide the labeling. To alleviate the problem of time-consuming labeling process, some methods have been suggested such as weak-supervised method, one-shot learning, self-supervised, suggestive learning, and so on. In this manuscript, those methods are analyzed and its possible future direction of the research is suggested.

Audio Event Detection Based on Attention CRNN (Attention CRNN에 기반한 오디오 이벤트 검출)

  • Kwak, Jin-Yeol;Chung, Yong-Joo
    • The Journal of the Korea institute of electronic communication sciences
    • /
    • v.15 no.3
    • /
    • pp.465-472
    • /
    • 2020
  • Recently, various deep neural networks based methods have been proposed for audio event detection. In this study, we improved the performance of audio event detection by adopting an attention approach to a baseline CRNN. We applied context gating at the input of the baseline CRNN and added an attention layer at the output. We improved the performance of the attention based CRNN by using the audio data of strong labels in frame units as well as the data of weak labels in clip levels. In the audio event detection experiments using the audio data from the Task 4 of the DCASE 2018/2019 Challenge, we could obtain maximally a 66% relative increase in the F-score in the proposed attention based CRNN compared with the baseline CRNN.

Text Region Detection Method Using Table Border Pseudo Label (표의 테두리 유사 라벨을 활용한 문자 영역 검출 방법)

  • Han, Jeong Hoon;Park, Se Jin;Moon, Young Shik
    • Journal of the Korea Institute of Information and Communication Engineering
    • /
    • v.24 no.10
    • /
    • pp.1271-1279
    • /
    • 2020
  • Text region detection is a technology that detects text area in handwriting or printed documents. The detected text areas are digitized through a recognition step, which is used in various fields depending on the purpose of use. However, the detection result of the small text unit is not suitable for the industrial field. In addition, the border of tables in the document that it causes miss-detected results, which has an adverse effect on the recognition step. To solve the issues, we propose a method for detecting text region using the border information of the table. In order to utilize the border information of the table, the proposed method adjusts the flow of two decoders. Experimentally, we show improved performance using the table border pseudo label based on weak supervised learning.

Dilated convolution and gated linear unit based sound event detection and tagging algorithm using weak label (약한 레이블을 이용한 확장 합성곱 신경망과 게이트 선형 유닛 기반 음향 이벤트 검출 및 태깅 알고리즘)

  • Park, Chungho;Kim, Donghyun;Ko, Hanseok
    • The Journal of the Acoustical Society of Korea
    • /
    • v.39 no.5
    • /
    • pp.414-423
    • /
    • 2020
  • In this paper, we propose a Dilated Convolution Gate Linear Unit (DCGLU) to mitigate the lack of sparsity and small receptive field problems caused by the segmentation map extraction process in sound event detection with weak labels. In the advent of deep learning framework, segmentation map extraction approaches have shown improved performance in noisy environments. However, these methods are forced to maintain the size of the feature map to extract the segmentation map as the model would be constructed without a pooling operation. As a result, the performance of these methods is deteriorated with a lack of sparsity and a small receptive field. To mitigate these problems, we utilize GLU to control the flow of information and Dilated Convolutional Neural Networks (DCNNs) to increase the receptive field without additional learning parameters. For the performance evaluation, we employ a URBAN-SED and self-organized bird sound dataset. The relevant experiments show that our proposed DCGLU model outperforms over other baselines. In particular, our method is shown to exhibit robustness against nature sound noises with three Signal to Noise Ratio (SNR) levels (20 dB, 10 dB and 0 dB).

Antecedents of Manufacturer's Private Label Program Engagement : A Focus on Strategic Market Management Perspective (제조업체 Private Labels 도입의 선행요인 : 전략적 시장관리 관점을 중심으로)

  • Lim, Chae-Un;Yi, Ho-Taek
    • Journal of Distribution Research
    • /
    • v.17 no.1
    • /
    • pp.65-86
    • /
    • 2012
  • The $20^{th}$ century was the era of manufacturer brands which built higher brand equity for consumers. Consumers moved from generic products of inconsistent quality produced by local factories in the $19^{th}$ century to branded products from global manufacturers and manufacturer brands reached consumers through distributors and retailers. Retailers were relatively small compared to their largest suppliers. However, sometime in the 1970s, things began to slowly change as retailers started to develop their own national chains and began international expansion, and consolidation of the retail industry from mom-and-pop stores to global players was well under way (Kumar and Steenkamp 2007, p.2) In South Korea, since the middle of the 1990s, the bulking up of retailers that started then has changed the balance of power between manufacturers and retailers. Retailer private labels, generally referred to as own labels, store brands, distributors own private-label, home brand or own label brand have also been performing strongly in every single local market (Bushman 1993; De Wulf et al. 2005). Private labels now account for one out of every five items sold every day in U.S. supermarkets, drug chains, and mass merchandisers (Kumar and Steenkamp 2007), and the market share in Western Europe is even larger (Euromonitor 2007). In the UK, grocery market share of private labels grew from 39% of sales in 2008 to 41% in 2010 (Marian 2010). Planet Retail (2007, p.1) recently concluded that "[PLs] are set for accelerated growth, with the majority of the world's leading grocers increasing their own label penetration." Private labels have gained wide attention both in the academic literature and popular business press and there is a glowing academic research to the perspective of manufacturers and retailers. Empirical research on private labels has mainly studies the factors explaining private labels market shares across product categories and/or retail chains (Dahr and Hoch 1997; Hoch and Banerji, 1993), factors influencing the private labels proneness of consumers (Baltas and Doyle 1998; Burton et al. 1998; Richardson et al. 1996) and factors how to react brand manufacturers towards PLs (Dunne and Narasimhan 1999; Hoch 1996; Quelch and Harding 1996; Verhoef et al. 2000). Nevertheless, empirical research on factors influencing the production in terms of a manufacturer-retailer is rather anecdotal than theory-based. The objective of this paper is to bridge the gap in these two types of research and explore the factors which influence on manufacturer's private label production based on two competing theories: S-C-P (Structure - Conduct - Performance) paradigm and resource-based theory. In order to do so, the authors used in-depth interview with marketing managers, reviewed retail press and research and presents the conceptual framework that integrates the major determinants of private labels production. From a manufacturer's perspective, supplying private labels often starts on a strategic basis. When a manufacturer engages in private labels, the manufacturer does not have to spend on advertising, retailer promotions or maintain a dedicated sales force. Moreover, if a manufacturer has weak marketing capabilities, the manufacturer can make use of retailer's marketing capability to produce private labels and lessen its marketing cost and increases its profit margin. Figure 1. is the theoretical framework based on a strategic market management perspective, integrated concept of both S-C-P paradigm and resource-based theory. The model includes one mediate variable, marketing capabilities, and the other moderate variable, competitive intensity. Manufacturer's national brand reputation, firm's marketing investment, and product portfolio, which are hypothesized to positively affected manufacturer's marketing capabilities. Then, marketing capabilities has negatively effected on private label production. Moderating effects of competitive intensity are hypothesized on the relationship between marketing capabilities and private label production. To verify the proposed research model and hypotheses, data were collected from 192 manufacturers (212 responses) who are producing private labels in South Korea. Cronbach's alpha test, explanatory / comfirmatory factor analysis, and correlation analysis were employed to validate hypotheses. The following results were drawing using structural equation modeling and all hypotheses are supported. Findings indicate that manufacturer's private label production is strongly related to its marketing capabilities. Consumer marketing capabilities, in turn, is directly connected with the 3 strategic factors (e.g., marketing investment, manufacturer's national brand reputation, and product portfolio). It is moderated by competitive intensity between marketing capabilities and private label production. In conclusion, this research may be the first study to investigate the reasons manufacturers engage in private labels based on two competing theoretic views, S-C-P paradigm and resource-based theory. The private label phenomenon has received growing attention by marketing scholars. In many industries, private labels represent formidable competition to manufacturer brands and manufacturers have a dilemma with selling to as well as competing with their retailers. The current study suggests key factors when manufacturers consider engaging in private label production.

  • PDF

Immunohistochemical Localization of Anoctamin 1 in the Mouse Cerebellum

  • Park, Yong Soo;Jeon, Ji Hyun;Lee, Seung Hee;Paik, Sun Sook;Kim, In-Beom
    • Applied Microscopy
    • /
    • v.48 no.4
    • /
    • pp.110-116
    • /
    • 2018
  • Since a transmembrane protein, TMEM16A, also called anoctamin 1 (ANO1), was identified as a bona fide calcium ($Ca^{2+}$)-activated chloride ($Cl^-$) channel (CaCC), there have been many reports on its expression and function. However, limited information on ANO1 expression and function in the brain is still available. In this study, we tried to reexamine expression patterns of ANO1 in the mouse cerebellum and further characterize ANO1-expressing components by immunohistochemical analyses. Strong ANO1 immunoreactivity was observed as large puncta in the granule cell layer and weak to moderate immunoreactivities were observed as small puncta in the molecular and Purkinje cell layers. Double-label experiments revealed that ANO1 did not colocalize with cerebellar neuronal population markers, such as anti-calbindin and anti-NeuN, while it colocalized or intermingled with a presynaptic marker, anti-synaptophysin. These results demonstrate that ANO1 is mainly localized at presynaptic terminals in the cerebellum and involved in synaptic transmission and modulation in cerebellar information processing.

Fusion System of Time-of-Flight Sensor and Stereo Cameras Considering Single Photon Avalanche Diode and Convolutional Neural Network (SPAD과 CNN의 특성을 반영한 ToF 센서와 스테레오 카메라 융합 시스템)

  • Kim, Dong Yeop;Lee, Jae Min;Jun, Sewoong
    • The Journal of Korea Robotics Society
    • /
    • v.13 no.4
    • /
    • pp.230-236
    • /
    • 2018
  • 3D depth perception has played an important role in robotics, and many sensory methods have also proposed for it. As a photodetector for 3D sensing, single photon avalanche diode (SPAD) is suggested due to sensitivity and accuracy. We have researched for applying a SPAD chip in our fusion system of time-of-fight (ToF) sensor and stereo camera. Our goal is to upsample of SPAD resolution using RGB stereo camera. Currently, we have 64 x 32 resolution SPAD ToF Sensor, even though there are higher resolution depth sensors such as Kinect V2 and Cube-Eye. This may be a weak point of our system, however we exploit this gap using a transition of idea. A convolution neural network (CNN) is designed to upsample our low resolution depth map using the data of the higher resolution depth as label data. Then, the upsampled depth data using CNN and stereo camera depth data are fused using semi-global matching (SGM) algorithm. We proposed simplified fusion method created for the embedded system.

No-Reference Image Quality Assessment based on Quality Awareness Feature and Multi-task Training

  • Lai, Lijing;Chu, Jun;Leng, Lu
    • Journal of Multimedia Information System
    • /
    • v.9 no.2
    • /
    • pp.75-86
    • /
    • 2022
  • The existing image quality assessment (IQA) datasets have a small number of samples. Some methods based on transfer learning or data augmentation cannot make good use of image quality-related features. A No Reference (NR)-IQA method based on multi-task training and quality awareness is proposed. First, single or multiple distortion types and levels are imposed on the original image, and different strategies are used to augment different types of distortion datasets. With the idea of weak supervision, we use the Full Reference (FR)-IQA methods to obtain the pseudo-score label of the generated image. Then, we combine the classification information of the distortion type, level, and the information of the image quality score. The ResNet50 network is trained in the pre-train stage on the augmented dataset to obtain more quality-aware pre-training weights. Finally, the fine-tuning stage training is performed on the target IQA dataset using the quality-aware weights to predicate the final prediction score. Various experiments designed on the synthetic distortions and authentic distortions datasets (LIVE, CSIQ, TID2013, LIVEC, KonIQ-10K) prove that the proposed method can utilize the image quality-related features better than the method using only single-task training. The extracted quality-aware features improve the accuracy of the model.

Research on Community Knowledge Modeling of Readers Based on Interest Labels

  • Kai, Wang;Wei, Pan;Xingzhi, Chen
    • Journal of Information Processing Systems
    • /
    • v.19 no.1
    • /
    • pp.55-66
    • /
    • 2023
  • Community portraits can deeply explore the characteristics of community structures and describe the personalized knowledge needs of community users, which is of great practical significance for improving community recommendation services, as well as the accuracy of resource push. The current community portraits generally have the problems of weak perception of interest characteristics and low degree of integration of topic information. To resolve this problem, the reader community portrait method based on the thematic and timeliness characteristics of interest labels (UIT) is proposed. First, community opinion leaders are identified based on multi-feature calculations, and then the topic features of their texts are identified based on the LDA topic model. On this basis, a semantic mapping including "reader community-opinion leader-text content" was established. Second, the readers' interest similarity of the labels was dynamically updated, and two kinds of tag parameters were integrated, namely, the intensity of interest labels and the stability of interest labels. Finally, the similarity distance between the opinion leader and the topic of interest was calculated to obtain the dynamic interest set of the opinion leaders. Experimental analysis was conducted on real data from the Douban reading community. The experimental results show that the UIT has the highest average F value (0.551) compared to the state-of-the-art approaches, which indicates that the UIT has better performance in the smooth time dimension.

A Review of SERS for Biomaterials Analysis Using Metal Nanoparticles (바이오 물질 분석을 위한 금속 나노입자를 이용한 SERS 분석 연구동향)

  • Jang, Eue-Soon
    • Ceramist
    • /
    • v.22 no.3
    • /
    • pp.281-300
    • /
    • 2019
  • Surface enhanced Raman scattering (SERS) was first discovered in 1974 by an unexpected Raman signal increase from Pyridine adsorbed on rough Ag electrode surfaces by the M. Fleishmann group. M. Moskovits group suggested that this phenomenon could be caused by surface plasmon resonance (SPR), which is a collective oscillation of free electrons at the surface of metal nanostructures by an external light source. After about 40 years, the SERS study has attracted great attention as a biomolecule analysis technology, and more than 2500 new papers and 500 review papers related to SERS topic have been published each year in recently. The advantages of biomaterials analysis using SERS are as follows; ① Molecular level analysis is possible based on unique fingerprint information of biomolecule, ② There is no photo-bleaching effect of the Raman reporters, allowing long-term monitoring of biomaterials compared to fluorescence microscopy, ③ SERS peak bandwidth is approximately 10 to 100 times narrower than fluorescence emission from organic phosphor or quantum dot, resulting in higher analysis accuracy, ④ Single excitation wavelength allows analysis of various biomaterials, ⑤ By utilizing near-infrared (NIR) SERS-activated nanostructures and NIR excitation lasers, auto-fluorescence noise in the visible wavelength range can be avoided from in vivo experiment and light damage in living cells can be minimized compared to visible lasers, ⑥ The weak Raman signal of the water molecule makes it easy to analyze biomaterials in aqueous solutions. For this reason, SERS is attracting attention as a next-generation non-invasive medical diagnostic device as well as substance analysis. In this review, the principles of SERS and various biomaterial analysis principles using SERS analysis will be introduced through recent research papers.