• Proceedings of the Korean Society of Computer Information Conference
• /
• 2020.07a
• /
• pp.13-14
• /
• 2020

본 논문에서는 엣지 컴퓨팅 환경에서 머신러닝을 활용해 에스컬레이터 이상 감지 및 결함 분류를 하는 연구를 진행하였다. 엣지 컴퓨팅 기반 머신러닝을 사용해 에스컬레이터의 이상 감지 및 결함 분류를 위한 OneM2M환경을 구축하였으며 에스컬레이터에서 발생하는 소음에서 고장 유형에 따라 나타나는 주파수를 이용한다. Edge TPU를 활용해 엣지 컴퓨팅 시스템의 처리량을 최대화하고, 각 작업의 수행시간을 최소화함으로써 엣지 컴퓨팅 환경에서 이상 감지와 결함 분류를 수행할 수 있다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.435-437
• /
• 2022

보행 보조 로봇의 효과적인 보조를 위해서는 사용자의 보행 유형을 인식하는 것이 중요하다. 본 논문에서는 end-to-end 분류 네트워크 기반 보행 환경 인식 방법을 사용하여 사용자의 보행 유형을 강인하게 추정한다. 실외 보행 환경을 오르막길, 평지, 내리막길 3 가지로 분류하는 딥러닝 모델을 학습시켰으며, 엣지 디바이스에서 이를 사용하기 위해 네트워크 경량화를 진행하였다. 경량화 후 추론 속도는 약 47FPS 수준으로 실시간으로 보행 보조 로봇에 적용 가능한 것을 검증했으며, 정확도 측면에서도 97% 이상의 성능을 얻을 수 있었다.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.10
• /
• pp.67-76
• /
• 2023

Although deep learning models are making innovative achievements in the field of computer vision, the problem of vulnerability to adversarial examples continues to be raised. Adversarial examples are attack methods that inject fine noise into images to induce misclassification, which can pose a serious threat to the application of deep learning models in the real world. In this paper, we propose a model that detects adversarial examples using differences in predictive values between edge-learned classification models and underlying classification models. The simple process of extracting the edges of the objects and reflecting them in learning can increase the robustness of the classification model, and economical and efficient detection is possible by detecting adversarial examples through differences in predictions between models. In our experiments, the general model showed accuracy of {49.9%, 29.84%, 18.46%, 4.95%, 3.36%} for adversarial examples (eps={0.02, 0.05, 0.1, 0.2, 0.3}), whereas the Canny edge model showed accuracy of {82.58%, 65.96%, 46.71%, 24.94%, 13.41%} and other edge models showed a similar level of accuracy also, indicating that the edge model was more robust against adversarial examples. In addition, adversarial example detection using differences in predictions between models revealed detection rates of {85.47%, 84.64%, 91.44%, 95.47%, and 87.61%} for each epsilon-specific adversarial example. It is expected that this study will contribute to improving the reliability of deep learning models in related research and application industries such as medical, autonomous driving, security, and national defense.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.539-540
• /
• 2023

최근 물체 인식 모델의 성능을 개선하기 위한 다양한 연구가 진행 중이다. 본 논문에서는 K-means 기반 앵커박스 선정 기법을 적용한 새로운 물체 인식 모델 성능 개선 방법을 제안한다. 제안된 방법은 항만 내 설치된 컨테이너 사고를 예방하기 위한 컨테이너 사고위험도 분류 모델에 적용하여 성능 평가를 하였다. 특히, 컨테이너 사고위험도 분류 모델은 작은 물체를 인식해야 하며 이런 환경에서는 기존 물체 인식 모델 성능이 낮게 나타난다. 본 논문에서는 제안한 K-means 기반 앵커박스 선정 기법을 적용하여 물체 인식 모델 성능이 개선됨을 확인하였디.

• Transactions on Semiconductor Engineering
• /
• v.2 no.1
• /
• pp.17-20
• /
• 2024

This paper presents a novel lightweight object detection model tailored for low-powered edge devices, addressing the limitations of traditional resource-intensive computer vision models. Our proposed detector, inspired by the Single Shot Detector (SSD), employs a compact yet robust network design. Crucially, it integrates an 'enhancer block' that significantly boosts its efficiency in detecting smaller objects. The model comprises two primary components: the Light_Block for efficient feature extraction using Depth-wise and Pointwise Convolution layers, and the Enhancer_Block for enhanced detection of tiny objects. Trained from scratch on the Udacity Annotated Dataset with image dimensions of 300x480, our model eschews the need for pre-trained classification weights. Weighing only 5.5MB with approximately 0.43M parameters, our detector achieved a mean average precision (mAP) of 27.7% and processed at 140 FPS, outperforming conventional models in both precision and efficiency. This research underscores the potential of lightweight designs in advancing object detection for edge devices without compromising accuracy.

• Journal of Internet of Things and Convergence
• /
• v.10 no.2
• /
• pp.17-23
• /
• 2024

Losses in domestic water supply due to leaks are very large, such as fractures and defects in pipelines. Therefore, preventive measures to prevent water leakage are necessary. We propose the development of a leakage detection sensor utilizing vibration sensors and present an optimal leakage detection algorithm leveraging artificial intelligence. Vibrational sound data acquired from water pipelines undergo a preprocessing stage using FFT (Fast Fourier Transform), followed by leakage classification using an optimized tree-based boosting algorithm. Applying this method to approximately 260,000 experimental data points from various real-world scenarios resulted in a 97% accuracy, a 4% improvement over existing SVM(Support Vector Machine) methods. The processing speed also increased approximately 80 times, confirming its suitability for edge device applications.

• Journal of the Institute of Convergence Signal Processing
• /
• v.23 no.2
• /
• pp.107-114
• /
• 2022

In order to classify 3D film images as right or wrong, it is necessary to detect the pattern in a 3D film image. However, if the contrast of the pixels in the 3D film image is low, it is not easy to classify as the right and wrong 3D film images because the pattern in the image might not be clear. In this paper, we propose a method of classifying 3D film images as right or wrong by comparing the width at a specific frequency of each histogram after obtaining the histogram. Since, it is classified using the width of the histogram, the analysis process is not complicated. From the experiment, the histograms of right and wrong 3D film images were distinctly different, and the proposed algorithm reflects these features, and showed that all 3D film images were accurately classified at a specific frequency of the histogram. The performance of the proposed algorithm was verified to be the best through the comparison test with the other methods such as image subtraction, otsu thresholding, canny edge detection, morphological geodesic active contour, and support vector machines, and it was shown that excellent classification accuracy could be obtained without detecting the patterns in 3D film images.