• IEMEK Journal of Embedded Systems and Applications
• /
• v.15 no.5
• /
• pp.227-234
• /
• 2020

In this paper, we propose wav-U-Net to improve speech enhancement in heavy noisy environments, and it has implemented three principal techniques. First, as input data, we use 128 modified Mel-scale filter banks which can reduce computational burden instead of 512 frequency bins. Mel-scale aims to mimic the non-linear human ear perception of sound by being more discriminative at lower frequencies and less discriminative at higher frequencies. Therefore, Mel-scale is the suitable feature considering both performance and computing power because our proposed network focuses on speech signals. Second, we add a simple ResNet as pre-processing that helps our proposed network make estimated speech signals clear and suppress high-frequency noises. Finally, the proposed U-Net model shows significant performance regardless of the kinds of noise. Especially, despite using a single channel, we confirmed that it can well deal with non-stationary noises whose frequency properties are dynamically changed, and it is possible to estimate speech signals from noisy speech signals even in extremely noisy environments where noises are much lauder than speech (less than SNR 0dB). The performance on our proposed wav-U-Net was improved by about 200% on SDR and 460% on NSDR compared to the conventional Jansson's wav-U-Net. Also, it was confirmed that the processing time of out wav-U-Net with 128 modified Mel-scale filter banks was about 2.7 times faster than the common wav-U-Net with 512 frequency bins as input values.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.391-394
• /
• 2023

일반적으로 화재 감지 시스템은 정확하고 빠르게 화재를 감지하는 것은 어려운 문제 중 하나이다. 본 논문에서는 U-net을 활용하여 기존의 화재(불) 영역 추출 기법으로 Segmentation으로 보다 정밀하게 탐지하는 기법을 제안한다. 이 기법은 화재 이미지에서 연기제거 및 색상보정을 통해 이미지를 전처리하여 화염 영역을 추출한 뒤 U-Net으로 학습시켜 이미지를 입력하면 불 영역의 Segmentation을 추출하도록 한다.

• Computers and Concrete
• /
• v.32 no.6
• /
• pp.615-623
• /
• 2023

Crack detection is an essential method to ensure the safety of dam concrete structures. Low-quality crack images of dam concrete structures limit the application of neural network methods in crack detection. This research proposes a modified attentional mechanism model to reduce the disturbance caused by uneven light, shadow, and water spots in crack images. Also, the focal loss function solves the small ratio of crack information. The dataset collects from the network, laboratory and actual inspection dataset of dam concrete structures. This research proposes a novel method for crack detection of dam concrete structures based on the U-Net neural network, namely AF-UNet. A mutual comparison of OTSU, Canny, region growing, DeepLab V3+, SegFormer, U-Net, and AF-UNet (proposed) verified the detection accuracy. A binocular camera detects cracks in the experimental scene. The smallest measurement width of the system is 0.27 mm. The potential goal is to achieve real-time detection and localization of cracks in dam concrete structures.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.81-83
• /
• 2021

In this paper, we propose a method to segment a plant from a plant image using U-Net. The network is an end-to-end fully convolutional network that is mainly used for image segmentation. When training the network, we used a binary image that is acquired by the manual segmentation of a plant from the background. Experimental results show that the U-Net based segmentation network can extract a plant from a digital image accurately.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.221-235
• /
• 2022

Manual inspection of steel box girders on long span bridges is time-consuming and labor-intensive. The quality of inspection relies on the subjective judgements of the inspectors. This study proposes an automated approach to detect and segment cracks in high-resolution images. An end-to-end cascaded framework is proposed to first detect the existence of cracks using a deep convolutional neural network (CNN) and then segment the crack using a modified U-Net encoder-decoder architecture. A Naïve Bayes data fusion scheme is proposed to reduce the false positives and false negatives effectively. To generate the binary crack mask, first, the original images are divided into 448 × 448 overlapping image patches where these image patches are classified as cracks versus non-cracks using a deep CNN. Next, a modified U-Net is trained from scratch using only the crack patches for segmentation. A customized loss function that consists of binary cross entropy loss and the Dice loss is introduced to enhance the segmentation performance. Additionally, a Naïve Bayes fusion strategy is employed to integrate the crack score maps from different overlapping crack patches and to decide whether a pixel is crack or not. Comprehensive experiments have demonstrated that the proposed approach achieves an 81.71% mean intersection over union (mIoU) score across 5 different training/test splits, which is 7.29% higher than the baseline reference implemented with the original U-Net.

• Smart Media Journal
• /
• v.11 no.7
• /
• pp.94-103
• /
• 2022

Weeds bring disadvantages to crops since they can damage them, and a clean treatment with less pollution and contamination should be developed. Artificial intelligence gives new hope to agriculture to achieve smart farming. This study delivers an automated weeds growth point detection using deep learning. This study proposes a combination of semantic graphics for generating data annotation and U-Net with pre-trained deep learning as a backbone for locating the growth point of the weeds on the given field scene. The dataset was collected from an actual field. We measured the intersection over union, f1-score, precision, and recall to evaluate our method. Moreover, Mobilenet V2 was chosen as the backbone and compared with Resnet 34. The results showed that the proposed method was accurate enough to detect the growth point and handle the brightness variation. The best performance was achieved by Mobilenet V2 as a backbone with IoU 96.81%, precision 97.77%, recall 98.97%, and f1-score 97.30%.

• Journal of Biomedical Engineering Research
• /
• v.42 no.3
• /
• pp.100-106
• /
• 2021

Measuring Intima-media thickness (IMT) with ultrasound images can help early detection of coronary artery disease. As a result, numerous machine learning studies have been conducted to measure IMT. However, most of these studies require several steps of pre-treatment to extract the boundary, and some require manual intervention, so they are not suitable for on-site treatment in urgent situations. in this paper, we propose to use deep learning networks U-Net, Attention U-Net, and Pretrained U-Net to automatically segment the intima-media complex. This study also applied the HE, HS, and CLAHE preprocessing technique to wireless portable ultrasound diagnostic device images. As a result, The average dice coefficient of HE applied Models is 71% and CLAHE applied Models is 70%, while the HS applied Models have improved as 72% dice coefficient. Among them, Pretrained U-Net showed the highest performance with an average of 74%. When comparing this with the mean value of IMT measured by Conventional wired ultrasound equipment, the highest correlation coefficient value was shown in the HS applied pretrained U-Net.

• Journal of Korea Multimedia Society
• /
• v.21 no.12
• /
• pp.1363-1369
• /
• 2018

Breast cancer is one of the most common cancers in women worldwide. In Korea, breast cancer is most common cancer in women followed by thyroid cancer. The purpose of this study is to evaluate the possibility of using deep - run model for segmentation of breast masses and to identify the best deep-run model for breast mass segmentation. In this study, data of patients with breast masses were collected at Asan Medical Center. We used 596 images of mammography and 596 images of gold standard. In the area of interest of the medical image, it was cut into a rectangular shape with a margin of about 10% up and down, and then converted into an 8-bit image by adjusting the window width and level. Also, the size of the image was resampled to $150{\times}150$. In Deconvolution net, the average accuracy is 91.78%. In U-net, the average accuracy is 90.09%. Deconvolution net showed slightly better performance than U-net in this study, so it is expected that deconvolution net will be better for breast mass segmentation. However, because of few cases, there are a few images that are not accurately segmented. Therefore, more research is needed with various training data.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.01a
• /
• pp.65-68
• /
• 2021

본 연구에서는 Dilated U-Net 기반의 이미지 복원기법을 통해 콘크리트 균열 추출 성능 개선 방안을 제안한다. 콘크리트 균열은 구조물의 미관상의 문제뿐 아니라 추후 큰 안전사고의 원인이 될 수 있어 초기대응이 중요하다. 현재는 점검자가 직접 육안으로 검사하는 외관 검사법이 주로 사용되고 있지만, 이는 정확성 및 비용, 시간, 그리고 안전성 면에서 한계를 갖고 있다. 이에 콘크리트 구조물 표면에 대해 획득한 영상 처리 기법을 사용한 검사 방식 도입의 관심이 늘어나고 있다. 또한, 딥러닝 기술의 발달로 딥러닝을 적용한 영상처리의 연구 역시 활발하게 진행되고 있다. 본 연구는 콘크리트 균열 추개선출 성능 개선을 위해 Dilated U-Net 기반의 이미지 복원기법을 적용하는 방안을 제안하였고 성능 검증 결과, 기존 U-Net 기반의 정확도가 98.78%, 조화평균 82.67%였던 것에 비해 정확도 99.199%, 조화평균 88.722%로 성능이 되었음을 확인하였다.

• Journal of Korea Multimedia Society
• /
• v.24 no.7
• /
• pp.849-859
• /
• 2021

Prostate cancer (PCa) is a fatal disease that occurs in men. In general, PCa cells are found in the prostate gland. Early diagnosis is the key to prevent the spreading of cancers to other parts of the body. In this case, deep learning-based systems can detect and distinguish histological patterns in microscopy images. The histological grades used for the analysis were benign, grade 3, grade 4, and grade 5. In this study, we attempt to use transfer learning and fine-tuning methods as well as different model architectures to develop and compare the models. We implemented MobileNet, ResNet50, and DenseNet121 models and used three different strategies of freezing layers techniques of fine-tuning, to get various pre-trained weights to improve accuracy. Finally, transfer learning using MobileNet with the half-layer frozen showed the best results among the nine models, and 90% accuracy was obtained on the test data set.