• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.3
• /
• pp.89-100
• /
• 2023

Recently, on-device artificial intelligence (AI) solutions using mobile devices and embedded edge devices have emerged in various fields, such as computer vision, to address network traffic burdens, low-energy operations, and security problems. Although vision transformer deep learning models have outperformed conventional convolutional neural network (CNN) models in computer vision, they require more computations and parameters than CNN models. Thus, they are not directly applicable to embedded edge devices with limited hardware resources. Many researchers have proposed various model compression methods or lightweight architectures for vision transformers; however, there are only a few studies evaluating the effects of model compression techniques of vision transformers on performance. Regarding this problem, this paper presents a performance evaluation of vision transformers on embedded platforms. We investigated the behaviors of three vision transformers: DeiT, LeViT, and MobileViT. Each model performance was evaluated by accuracy and inference time on edge devices using the ImageNet dataset. We assessed the effects of the quantization method applied to the models on latency enhancement and accuracy degradation by profiling the proportion of response time occupied by major operations. In addition, we evaluated the performance of each model on GPU and EdgeTPU-based edge devices. In our experimental results, LeViT showed the best performance in CPU-based edge devices, and DeiT-small showed the highest performance improvement in GPU-based edge devices. In addition, only MobileViT models showed performance improvement on EdgeTPU. Summarizing the analysis results through profiling, the degree of performance improvement of each vision transformer model was highly dependent on the proportion of parts that could be optimized in the target edge device. In summary, to apply vision transformers to on-device AI solutions, either proper operation composition and optimizations specific to target edge devices must be considered.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.2-2
• /
• 2023

기후변화가 가속화로 인해 수재해의 빈도와 강도 예측이 어려워짐에 따라 실시간 홍수 모니터링에 대한 수요가 증가하고 있다. 합성개구레이다는 광원과 날씨에 무관하게 촬영이 가능하여 수재해 발생시에도 영상을 확보할 수 있다. 합성개구레이다를 활용한 수체 탐지 알고리즘 개발이 활발히 연구되어 왔고, 딥러닝의 발달로 CNN을 활용하여 높은 정확도로 수체 탐지가 기능해졌다. 하지만, CNN 기반 수체 탐지 모델은 훈련시 높은 정량적 정확성 지표를 달성하여도 추론 후 정성적 평가시 경계와 소하천에 대한 탐지 정확성이 떨어진다. 홍수 모니터링에서 특히 중요한 정보인 경계와 좁은 하천에 대해서 정확성이 떨어짐에 따라 실생활 적용이 어렵다. 이에 경계를 강화한 적대적 학습 기반의 수체 탐지 모델을 개발하여 더 세밀하고 정확하게 탐지하고자 한다. 적대적 학습은 생성적 적대 신경망(GAN)의 두 개의 모델인 생성자와 판별자가 서로 관여하며 더 높은 정확도를 달성할 수 있도록 학습이다. 이러한 적대적 학습 개념을 수체 탐지 모델에 처음으로 도입하여, 생성자는 실제 라벨 데이터와 유사하게 수체 경계와 소하천까지 탐지하고자 학습한다. 반면 판별자는 경계 거리 변환 맵과 합성개구레이다 영상을 기반으로 라벨데이터와 수체 탐지 결과를 구분한다. 경계가 강화될 수 있도록, 면적과 경계를 모두 고려할 수 있는 손실함수 조합을 구성하였다. 제안 모델이 경계와 소하천을 정확히 탐지하는지 판단하기 위해, 정량적 지표로 F1-score를 사용하였으며, 육안 판독을 통해 정성적 평가도 진행하였다. 기존 U-Net 모델이 탐지하지 못하던 영역에 대해 제안한 경계 강화 적대적 수체 탐지 모델이 수체의 세밀한 부분까지 탐지할 수 있음을 증명하였다.

• Journal of the Korea Society for Simulation
• /
• v.30 no.2
• /
• pp.41-48
• /
• 2021

As the rural population continues to decline and aging, the improvement of agricultural productivity is becoming more important. Early prediction of crop quality can play an important role in improving agricultural productivity and profitability. Although many researches have been conducted recently to classify diseases and predict crop yield using CNN based deep learning and transfer learning technology, there are few studies which predict postharvest crop quality early in the planting stage. In this study, a early quality prediction model is proposed for sprout ginseng, which is drawing attention as a healthy functional foods. For this end, we took pictures of ginseng seedlings in the planting stage and cultivated them through hydroponic cultivation. After harvest, quality data were labeled by classifying the quality of ginseng sprout. With this data, we build early quality prediction models using several pre-trained CNN models through transfer learning technology. And we compare the prediction performance such as learning period and accuracy between each model. The results show more than 80% prediction accuracy in all proposed models, especially ResNet152V2 based model shows the highest accuracy. Through this study, it is expected that it will be able to contribute to production and profitability by automating the existing seedling screening works, which primarily rely on manpower.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.6
• /
• pp.1137-1144
• /
• 2022

In this study, an artificial intelligence(AI) was developed to help with facial expression practice in order to express emotions. The developed AI used multimodal inputs consisting of sentences and facial images for deep neural networks (DNNs). The DNNs calculated similarities between the emotions predicted by the sentences and the emotions predicted by facial images. The user practiced facial expressions based on the situation given by sentences, and the AI provided the user with numerical feedback based on the similarity between the emotion predicted by sentence and the emotion predicted by facial expression. ResNet34 structure was trained on FER2013 public data to predict emotions from facial images. To predict emotions in sentences, KoBERT model was trained in transfer learning manner using the conversational speech dataset for emotion classification opened to the public by AIHub. The DNN that predicts emotions from the facial images demonstrated 65% accuracy, which is comparable to human emotional classification ability. The DNN that predicts emotions from the sentences achieved 90% accuracy. The performance of the developed AI was evaluated through experiments with changing facial expressions in which an ordinary person was participated.

• The Journal of the Convergence on Culture Technology
• /
• v.7 no.4
• /
• pp.675-681
• /
• 2021

With the recent growth of the AI speaker market, the demand for speech synthesis technology that enables natural conversation with users is increasing. Therefore, there is a need for a multi-speaker speech synthesis system that can generate voices of various tones. In order to synthesize natural speech, it is required to train with a large-capacity. high-quality speech DB. However, it is very difficult in terms of recording time and cost to collect a high-quality, large-capacity speech database uttered by many speakers. Therefore, it is necessary to train the speech synthesis system using the speech DB of a very large number of speakers with a small amount of training data for each speaker, and a technique for naturally expressing the tone and rhyme of multiple speakers is required. In this paper, we propose a technology for constructing a speaker encoder by applying the deep learning-based x-vector technique used in speaker recognition technology, and synthesizing a new speaker's tone with a small amount of data through the speaker encoder. In the multi-speaker speech synthesis system, the module for synthesizing mel-spectrogram from input text is composed of Tacotron2, and the vocoder generating synthesized speech consists of WaveNet with mixture of logistic distributions applied. The x-vector extracted from the trained speaker embedding neural networks is added to Tacotron2 as an input to express the desired speaker's tone.

• Journal of Broadcast Engineering
• /
• v.23 no.6
• /
• pp.855-865
• /
• 2018

Sound event detection is one of the research areas to model human auditory cognitive characteristics by recognizing events in an environment with multiple acoustic events and determining the onset and offset time for each event. DCASE, a research group on acoustic scene classification and sound event detection, is proceeding challenges to encourage participation of researchers and to activate sound event detection research. However, the size of the dataset provided by the DCASE Challenge is relatively small compared to ImageNet, which is a representative dataset for visual object recognition, and there are not many open sources for the acoustic dataset. In this study, the sound events that can occur in indoor and outdoor are collected on a larger scale and annotated for dataset construction. Furthermore, to improve the performance of the sound event detection task, we developed a dual CNN structured sound event detection system by adding a supplementary neural network to a convolutional neural network to determine the presence of sound events. Finally, we conducted a comparative experiment with both baseline systems of the DCASE 2016 and 2017.

• Journal of Industrial Convergence
• /
• v.21 no.5
• /
• pp.1-13
• /
• 2023

Carbon neutrality is the concept of reducing greenhouse gases emitted by human activities and making actual emissions zero through removal of remaining gases. It is also called "Net-Zero" and "carbon zero". Korea has declared a "2050 Carbon Neutrality policy" to cope with the climate change crisis. Various carbon reduction legislative processes are underway. Since carbon neutrality requires changes in industrial technology, it is important to prepare a system for carbon zero. This paper aims to understand the status and trends of global carbon neutrality technology. Therefore, ROK's web platform "www.naver.com." was selected as the data collection scope. Korean online articles related to carbon neutrality were collected. Carbon neutrality technology trends were analyzed by future signal methodology and Word2Vec algorithm which is a neural network deep learning technology. As a result, technology advancement in the steel and petrochemical sectors, which are carbon over-release industries, was required. Investment feasibility in the electric vehicle sector and technology advancement were on the rise. It seems that the government's support for carbon neutrality and the creation of global technology infrastructure should be supported. In addition, it is urgent to cultivate human resources, and possible to confirm the need to prepare support policies for carbon neutrality.