• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.160-162
• /
• 2019

음향 이벤트 인식 기술은 오디오 신호에서 음향 이벤트를 예측하는 기술로, 최근 대용량 데이터베이스의 배포, 인식 알고리즘과 하드웨어의 발전, 관련 인식 대회 등에 힘입어 많은 연구가 이루어지고 있는 분야이다. 본 논문에서는 음향 장면 및 이벤트 인식 관련 대회인 DCASE 챌린지에 대하여 기술하고, 약한 레이블 기반의 데이터를 학습해 강한 레이블을 예측하는 DCASE 챌린지 과제 4에 대하여 설명한다. 또한 DCASE 챌린지 과제 4에 제출된 다양한 음향 이벤트 인식 알고리즘과 데이터베이스의 종류에 따른 성능을 비교하여 음향 이벤트 인식 성능을 분석한다.

• Journal of the Korean Mathematical Society
• /
• v.59 no.5
• /
• pp.963-986
• /
• 2022

Given a dimension function 𝜔, we introduce the notion of an 𝜔-vector weighted digraph and an 𝜔-equivalence between them. Then we establish a bijection between the weakly (ℤ/2)ⁿ-equivariant homeomorphism classes of small covers over a product of simplices ∆^𝜔(1) × ⋯ × ∆^𝜔(m) and the set of 𝜔-equivalence classes of 𝜔-vector weighted digraphs with m-labeled vertices, where n is the sum of the dimensions of the simplicies. Using this bijection, we obtain a formula for the number of weakly (ℤ/2)ⁿ-equivariant homeomorphism classes of small covers over a product of three simplices.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.2
• /
• pp.139-147
• /
• 2021

This paper introduces and implements a Sound Event Detection (SED) system based on weakly-supervised learning where only part of the data is labeled, and analyzes the effect of parameters. The SED system estimates the classes and onset/offset times of events in the acoustic signal. In order to train the model, all information on the event class and onset/offset times must be provided. Unfortunately, the onset/offset times are hard to be labeled exactly. Therefore, in the weakly-supervised task, the SED model is trained by "strongly labeled data" including the event class and activations, "weakly labeled data" including the event class, and "unlabeled data" without any label. Recently, the SED systems using the mean-teacher model are widely used for the task with several parameters. These parameters should be chosen carefully because they may affect the performance. In this paper, performance analysis was performed on parameters, such as the feature, moving average parameter, weight of the consistency cost function, ramp-up length, and maximum learning rate, using the data of DCASE 2020 Task 4. Effects and the optimal values of the parameters were discussed.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.1
• /
• pp.24-31
• /
• 2020

In this paper, the deep convolutional neural network for sound event detection is studied. Especially, the end-to-end neural network, which generates the detection results from the input audio waveform, is studied for weakly supervised problem that includes weakly-labeled and unlabeled dataset. The proposed system is based on the network structure that consists of deeply-stacked 1-dimensional convolutional neural networks, and enhanced by the skip connection and gating mechanism. Additionally, the proposed system is enhanced by the sound event detection and post processings, and the training step using the mean-teacher model is added to deal with the weakly supervised data. The proposed system was evaluated by the Detection and Classification of Acoustic Scenes and Events (DCASE) 2019 Task 4 dataset, and the result shows that the proposed system has F₁-scores of 54 % (segment-based) and 32 % (event-based).

• Applied Microscopy
• /
• v.32 no.3
• /
• pp.247-255
• /
• 2002

In this study, we carried out immunostaining and immunogold labeling with antibodies to serotonin and somatostatin to examine the characteristics and functions of the neurons that secrete neurotransmitters in optic lobes of Todarodes pacificus and Octopus minor. As a result of immunostaining with anti-somatostatin, the nerve cells of Todarodes pacificus reacted as similar to the anti-serotonin, but in Octopus minor, only large cells in the outer granule cell layer reacted positively. In the immunogold labeling with anti-serotonin, the nerve cells in the inner grande cell layer and medulla of Todarodes pacificus reacted strongly, 30 gold particles being labeled per $0.5{\mu}m^2$ of the cytoplasm. However, in Octopus minor, only 17 gold particles were labeled, which stated a weak reaction. On the other hand, in the anti-somatostatin case, the nerve cells in the outer and inner granule cell layers and medulla of Todarodes pacificus showed strong reaction, 30 gold particles being labeled per $0.5{\mu}m^2$ of the cytoplasm while the nerve cells in the outer granule cell layer of Octopus minor reacted weakly, about 3 gold particles being labeled per the equivalent area. As a result of immunostaining and immunogold labeling with two types of antibodies to each part of the optic lobes, we found that the reactive nerve cells were distributed differently in the two species. In particular, the degree of reactivity to the immunostaining and immunogold labeling appeared stronger in Todarodes pacificus than in Octopus minor.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.3
• /
• pp.401-406
• /
• 2021

Recently, mean-teacher models based on convolutional recurrent neural networks are popularly used in audio event detection. The mean-teacher model is an architecture that consists of two parallel CRNNs and it is possible to train them effectively on the weakly-labelled and unlabeled audio data by using the consistency learning metric at the output of the two neural networks. In this study, we tried to improve the performance of the mean-teacher model by using additional derivative features of the log-mel spectrum. In the audio event detection experiments using the training and test data from the Task 4 of the DCASE 2018/2019 Challenges, we could obtain maximally a 8.1% relative decrease in the ER(Error Rate) in the mean-teacher model using proposed derivative features.

• Journal of Drive and Control
• /
• v.20 no.2
• /
• pp.15-23
• /
• 2023

In this study, two-dimensional location of crops for auto weeding was detected using deep learning. To construct a dataset for soybean detection, an image-capturing system was developed using a mono camera and single-board computer and the system was mounted on a weeding robot to collect soybean images. A dataset was constructed by extracting RoI (region of interest) from the raw image and each sample was labeled with soybean and the background for classification learning. The deep learning model consisted of four convolutional layers and was trained with a weakly supervised learning method that can provide object localization only using image-level labeling. Localization of the soybean area can be visualized via CAM and the two-dimensional position of the soybean was estimated by clustering the pixels associated with the soybean area and transforming the pixel coordinates to world coordinates. The actual position, which is determined manually as pixel coordinates in the image was evaluated and performances were 6.6(X-axis), 5.1(Y-axis) and 1.2(X-axis), 2.2(Y-axis) for MSE and RMSE about world coordinates, respectively. From the results, we confirmed that the center position of the soybean area derived through deep learning was sufficient for use in automatic weeding systems.

• Journal of Veterinary Clinics
• /
• v.36 no.4
• /
• pp.233-237
• /
• 2019

A 10-year-old castrated male Shih Tzu dog was submitted to a local animal hospital with a mass from gingiva to maxillofacial skeleton. Computed tomography revealed that strong invasion of the mass result in osteolysis in orbit and frontal bone. The excised mass was presented to the Pathology Department of the Veterinary Medicine, Jeju National University. Surgically excised mass was rubbery to firm in consistency. Histologically, the neoplastic mass was composed of irregular or interdigitating cords, islands or pseudo-glandular structures of stratified epithelial cells. These cords or islands showed typical palisading pattern of neoplastic epithelial cells to periphery without intercellular bridge (desmosome) and surrounded by eosinophilic immature collagenous matrix. Some area showed islands of well differentiated keratinizing squamous cell foci. Some lumen of glandular structures contained fibrin-like materials and RBC. These neoplastic cells showed marked invasive tendency to adjacent connective tissues and bony tissues, therefore solitary neoplastic cells were widely distributed throughout the surround connective tissue. The neoplastic cells showed positive reactions for pan-CK and CK14, weakly positive reaction for CK5/6. And the surrounding immature collagenous matrix was only labeled for vimentin.

• International Journal of Oral Biology
• /
• v.41 no.3
• /
• pp.133-139
• /
• 2016

The ultrastructural parameters related to synaptic release of endings which are presynaptic to tooth pulp afferent terminals (p-endings) were analyzed to understand the underlying mechanism for presynaptic modulation of tooth pulp afferents. Tooth pulp afferents were labelled by applying wheat-germ agglutinin conjugated horseradish peroxidase to the rat right lower incisor, whereafter electron microscopic morphometric analysis with serial section and reconstruction of p-endings in the trigeminal oral nucleus was performed. The results obtained from 15 p-endings presynaptic to 11 labeled tooth pulp afferent terminals were as follows. P-endings contained pleomorphic vesicles and made symmetrical synaptic contacts with labeled terminals. The p-endings showed small synaptic release-related ultrastructural parameters: volume, $0.82{\pm}0.45{\mu}m^3$ ($mean{\pm}SD$); surface area, $4.50{\pm}1.76{\mu}m^2$; mitochondrial volume, $0.15{\pm}0.07{\mu}m^3$; total apposed surface area, $0.69{\pm}0.24{\mu}m^2$; active zone area, $0.10{\pm}0.04{\mu}m^2$; total vesicle number, $1045{\pm}668.86$; and vesicle density, $1677{\pm}684/{\mu}m^2$. The volume of the p-endings showed strong positive correlation with the following parameters: surface area (r=0.97, P<0.01), mitochondrial volume (r=0.56, P<0.05), and total vesicle number (r=0.73, P<0.05). However, the volume of p-endings did not positively correlate or was very weakly correlated with the apposed surface area (r=-0.12, P=0.675) and active zone area (r=0.46, P=0.084). These results show that some synaptic release-related ultrastructural parameters of p-endings on the tooth pulp afferent terminals follow the "size principle" of Pierce and Mendell (1993) in the trigeminal nucleus oralis, but other parameters do not. Our findings may demonstrate a characteristic feature of synaptic release associated with p-endings.

• Korean Journal of Veterinary Research
• /
• v.37 no.2
• /
• pp.235-244
• /
• 1997

The present study was performed to investigate the patterns of appearance of lectin in trachea of fetuses of 60, 90 and 120 days old and neonates of Korean native goat. Carbohydrate markers were used in histochemistry for the determination of the lectin by staining of avidin-biotin-peroxidase complex(ABC), and the markers consisted of biotin-labeled concanavalin A(Con A), dolichos biflorus agglutinin(DBA), rincinus communis agglutinin(RCA-I), ulexeuropalus communis agglutinin(UEA) and wheat germ agglutinin(WGA). 1. The Con A-binding reactions appeared moderately on the apical surface of the tracheal epithelia in 60 days old fetuses, and the reactions were similar on the tracheal epithelia and glands in 90 and 120 days old fetuses and neonates. 2. Reaction of the DBA appeared as the strongest meanwhile the DBA-binding reactions were determined strongly on the apical surface of the tracheal epithelia in the 60 days old fetuses. Reaction for the DBA on the tracheal epithelia and glands of 90 and 120 days old fetuses and neonates were in same manner. 3. The RCA-I-binding reactions appeared very strongly on the apical surface of the tracheal epithelia in 60 and 90 days old fetuses. Reaction to the RCA-1 appeared moderately on certain apical surface of tracheal epithelia and glands in 120 days old fetuses and neonates. 4. No reactions provoked for the UEA in trachea of 60 days old fetuses and neonates, but the UEA-binding reactions appeared moderately in the tracheal epithelia of 90 days old fetuses and weakly in 120 days old fetuses. 5. The WGA-binding reactions appeared very strongly on the apical surface of the tracheal epithelia in 60 and 90 days old fetuses, and moreover, the reactions were determined on the luminal surface of the tracheal gland in 90 days old fetuses. On the other hand, goblet cells of the tracheal epithelia and glands in neonates reacted moderately to the WGA.