• Smart Media Journal
• /
• v.9 no.2
• /
• pp.92-98
• /
• 2020

This paper proposes a vehicle license plate recognition system using light weight deep learning models without high-end server. The proposed license plate recognition system consists of 3 steps: [license plate detection]-[character area segmentation]-[character recognition]. SSD-Mobilenet was used for license plate detection, ResNet with localization was used for character area segmentation, ResNet was used for character recognition. Experiemnts using Samsung Galaxy S7 and LG Q9, accuracy showed 85.3% accuracy and around 1.1 second running time.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.604-606
• /
• 2024

컴퓨터 비전 응용은 우리 생활에서 중요한 역할을 한다. 현재, 대규모 모델의 등장으로 딥 러닝의 훈련 및 운행 비용이 급격히 상승하고 있다. 자원이 제한된 환경에서는 일부 AI 프로그램을 실행할 수 없게 되므로, 경량화 연구가 필요하다. YOLOv8 은 현재 주요 목표 검출 모델 중 하나이며, 본 논문은 다중 스케일 Ghost 컨볼루션 모듈을 사용하여 구축된 새로운 YOLOv8-pose-msg 키포인트 검출 모델을 제안한다. 다양한 사양에서 새 모델의 매개변수 양은 최소 34% 감소할 수 있으며, 최대 59%까지 감소할 수 있다. 종합적인 검출 성능은 비교적 대규모 데이터셋에서 원래의 수준을 유지할 수 있으며, 소규모 데이터셋에서의 키포인트 검출은 30% 이상 증가할 수 있다. 동시에 최대 25%의 훈련 및 추론 시간을 절약할 수 있다.

• Journal of Digital Convergence
• /
• v.19 no.5
• /
• pp.307-313
• /
• 2021

In this paper, a lightweight network with fewer parameters compared to the existing object detection method is proposed. In the case of the currently used detection model, the network complexity has been greatly increased to improve accuracy. Therefore, the proposed network uses EfficientNet as a feature extraction network, and the subsequent layers are formed in a pyramid structure to utilize low-level detailed features and high-level semantic features. An attention process was applied between pyramid structures to suppress unnecessary noise for prediction. All computational processes of the network are replaced by depth-wise and point-wise convolutions to minimize the amount of computation. The proposed network was trained and evaluated using the PASCAL VOC dataset. The features fused through the experiment showed robust properties for various objects through a refinement process. Compared with the CNN-based detection model, detection accuracy is improved with a small amount of computation. It is considered necessary to adjust the anchor ratio according to the size of the object as a future study.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.10
• /
• pp.29-41
• /
• 2022

Recently, with the development of deep learning technology, a variety of huge models with excellent performance have been devised by pre-training massive amounts of text data. However, in order for such a model to be applied to real-life services, the inference speed must be fast and the amount of computation must be low, so the technology for model compression is attracting attention. Knowledge distillation, a representative model compression, is attracting attention as it can be used in a variety of ways as a method of transferring the knowledge already learned by the teacher model to a relatively small-sized student model. However, knowledge distillation has a limitation in that it is difficult to solve problems with low similarity to previously learned data because only knowledge necessary for solving a given problem is learned in a teacher model and knowledge distillation to a student model is performed from the same point of view. Therefore, we propose a heterogeneous knowledge distillation method in which the teacher model learns a higher-level concept rather than the knowledge required for the task that the student model needs to solve, and the teacher model distills this knowledge to the student model. In addition, through classification experiments on about 18,000 documents, we confirmed that the heterogeneous knowledge distillation method showed superior performance in all aspects of learning efficiency and accuracy compared to the traditional knowledge distillation.

• Journal of Convergence for Information Technology
• /
• v.10 no.8
• /
• pp.23-34
• /
• 2020

Recently, the artificial intelligence deep learning field has been hard to commercialize due to the high computing power and the price problem of computing resources. In this paper, we apply a double pruning techniques to evaluate the performance of the in-depth neural network and various datasets. Double pruning combines basic Network-slimming and Parameter-prunning. Our proposed technique has the advantage of reducing the parameters that are not important to the existing learning and improving the speed without compromising the learning accuracy. After training various datasets, the pruning ratio was increased to reduce the size of the model.We confirmed that MobileNet-V3 showed the highest performance as a result of NetScore performance analysis. We confirmed that the performance after pruning was the highest in MobileNet-V3 consisting of depthwise seperable convolution neural networks in the Cifar 10 dataset, and VGGNet and ResNet in traditional convolutional neural networks also increased significantly.

• Smart Media Journal
• /
• v.12 no.3
• /
• pp.49-59
• /
• 2023

Optical Camera Communication (OCC), known as the next-generation wireless communication technology, is currently under extensive research. The performance of OCC technology is affected by the communication environment, and various strategies are being studied to improve it. Among them, the most prominent method is applying convolutional neural networks (CNN) to the receiver of OCC using deep learning technology. However, in most studies, CNN is simply used to detect the transmitter. In this paper, we experiment with applying the convolutional neural network not only for transmitter detection but also for the Rx demodulation system. We hypothesize that, since the data images of the OCC system are relatively simple to classify compared to other image datasets, high accuracy results will appear in most CNN models. To prove this hypothesis, we designed and implemented an OCC system to collect data and applied it to 12 different CNN models for experimentation. The experimental results showed that not only high-performance CNN models with many parameters but also lightweight CNN models achieved an accuracy of over 99%. Through this, we confirmed the feasibility of applying the OCC system in real-time on mobile devices such as smartphones.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.11a
• /
• pp.42-43
• /
• 2022

본 논문은 모바일 단말에서 실시간으로 동작하는 딥러닝 기반 경량 초고해상화 기술에 관한 내용이다. 대용량 3차원 메쉬 모델의 비실시간 압축은 실시간 스트리밍 응용 시나리오에서 제약점으로 작용하고 있고, 본 논문에서는 두 입력 텐서의 차원을 일치시켜야 하는 element-wise 덧셈 대신 concatenation을 활용해 연산량을 개선하고, float-to-int8 양자화 과정에서 발생하는 오차를 줄이기 위해 weight clipping 및 regularization 기법을 활용해 초고해상화 화질 성능을 개선하였다. 제안하는 알고리즘은 기존 모바일 초고해상화 기술을 화질 측면에서 0.12dB, 처리 속도 측면에서 13.6ms 개선하였고, Mobile AI & AIM 2022 실시간 이미지 초고해상화 대회에서 1등을 수상하였다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.113-116
• /
• 2021

일반적으로 딥러닝 기반의 객체 검출(Object Detection)기법은 합성곱 신경망(Convolutional Neural Network, CNN)을 통해 입력된 영상의 특징(Feature)을 추출하여 이를 통해 객체 검출을 수행한다. 최근 자연어 처리 분야에서 획기적인 성능을 보인 트랜스포머(Transformer)가 영상 분류, 객체 검출과 같은 컴퓨터 비전 작업을 수행하는데 있어 경쟁력이 있음이 드러나고 있다. 본 논문에서는 YOLOv4-CSP의 CSP 블록을 개선한 one-stage 방식의 객체 검출 네트워크를 제안한다. 개선된 CSP 블록은 트랜스포머(Transformer)의 멀티 헤드 어텐션(Multi-Head Attention)과 CSP 형태의 공간 피라미드 풀링(Spatial Pyramid Pooling, SPP) 연산을 기반으로 네트워크의 Backbone과 Neck에서의 feature 학습을 돕는다. 본 실험은 MSCOCO test-dev2017 데이터 셋으로 평가하였으며 제안하는 네트워크는 YOLOv4-CSP의 경량화 모델인 YOLOv4s-mish에 대하여 평균 정밀도(Average Precision, AP)기준 2.7% 향상된 검출 정확도를 보인다.

• Journal of Industrial Convergence
• /
• v.20 no.12
• /
• pp.195-202
• /
• 2022

Underground Geospatial Information Map Management System(UGIMMS) integrates various underground facilities in the underground space into 3D mesh data, and supports to check the 3D image and location of the underground facilities in the mobile app. However, there is a problem that it takes a long time to run in the app because various underground facilities can exist in some areas executed by the app and can be seen layer by layer. In this paper, we propose a deep learning-based K-means vertex clustering algorithm as a method to reduce the execution time in the app by reducing the size of the data by reducing the number of vertices in the 3D mesh data within the range that does not cause a problem in visibility. First, our proposed method obtains refined vertex feature information through a deep learning encoder-decoder based model. And second, the method was simplified by grouping similar vertices through K-means vertex clustering using feature information. As a result of the experiment, when the vertices of various underground facilities were reduced by 30% with the proposed method, the 3D image model was slightly deformed, but there was no missing part, so there was no problem in checking it in the app.

• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.778-789
• /
• 2021

Recently, various lightweight methods for using Convolutional Neural Network(CNN) models in mobile devices have emerged. Weight quantization, which lowers bit precision of weights, is a lightweight method that enables a model to be used through integer calculation in a mobile environment where GPU acceleration is unable. Weight quantization has already been used in various models as a lightweight method to reduce computational complexity and model size with a small loss of accuracy. Considering the size of memory and computing speed as well as the storage size of the device and the limited network environment, this paper proposes a method of compressing integer weights after quantization using a video codec as a method. To verify the performance of the proposed method, experiments were conducted on VGG16, Resnet50, and Resnet18 models trained with ImageNet and Places365 datasets. As a result, loss of accuracy less than 2% and high compression efficiency were achieved in various models. In addition, as a result of comparison with similar compression methods, it was verified that the compression efficiency was more than doubled.