• 방송공학회논문지
• /
• 제27권3호
• /
• pp.283-294
• /
• 2022

딥 러닝 네트워크에서 사용되는 특징 맵은 일반적으로 영상보다 데이터가 크며 특징 맵을 전송하기 위해서는 영상의 압축률보다 더 높은 압축률이 요구된다. 본 논문은 딥러닝 기반의 영상처리에서 객체의 크기에 대한 강인성을 가지는 FPN 구조의 네트워크에서 사용되는 피라미드 특징 맵을 높은 압축률로 전송하기 위해 제안한 복원-예측 네트워크를 통해 전송된 일부 계층의 피라미드 특징 맵으로 전송하지 않은 계층의 피라미드 특징 맵을 예측하며, 압축으로 인한 손상을 복원하는 구조를 제안한다. 제안한 방법의 COCO 데이터셋 2017 Train images에 대한 객체 탐지의 성능은 rate-precision 그래프에서 VTM12.0을 통해 특징 맵을 압축한 결과 대비 BD-rate 31.25%의 성능향상을 보였고, PCA와 DeepCABAC을 통한 압축을 수행한 방법 대비 BD-rate 57.79%의 성능향상을 보였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• 반도체디스플레이기술학회지
• /
• 제22권3호
• /
• pp.149-154
• /
• 2023

This paper presents a novel approach to multi-focus image fusion using light field cameras. The proposed neural network, LFFCNN (Light Field Focus Convolutional Neural Network), is composed of three main modules: feature extraction, feature fusion, and feature reconstruction. Specifically, the feature extraction module incorporates SPP (Spatial Pyramid Pooling) to effectively handle images of various scales. Experimental results demonstrate that the proposed model not only effectively fuses a single All-in-Focus image from images with multi focus images but also offers more efficient and robust focus fusion compared to existing methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권3호
• /
• pp.701-720
• /
• 2023

Aerial objects are more challenging to segment than normal objects, which are usually smaller and have less textural detail. In the process of segmentation, target objects are easily omitted and misdetected, which is problematic. To alleviate these issues, we propose local aggregation feature pyramid networks (LAFPNs) and pyramid integrated context modules (PICMs) for aerial object segmentation. First, using an LAFPN, while strengthening the deep features, the extent to which low-level features interfere with high-level features is reduced, and numerous dense and small aerial targets are prevented from being mistakenly detected as a whole. Second, the PICM uses global information to guide local features, which enhances the network's comprehensive understanding of an entire image and reduces the missed detection of small aerial objects due to insufficient texture information. We evaluate our network with the MS COCO dataset using three categories: airplanes, birds, and kites. Compared with Mask R-CNN, our network achieves performance improvements of 1.7%, 4.9%, and 7.7% in terms of the AP metrics for the three categories. Without pretraining or any postprocessing, the segmentation performance of our network for aerial objects is superior to that of several recent methods based on classic algorithms.

• 한국정보통신학회논문지
• /
• 제25권4호
• /
• pp.523-529
• /
• 2021

4차산업 혁명 시대를 맞아, 기술의 변화가 다양한 분야에 적용되고 있다. 고지서 분야에서도 자동화, 디지털화, 데이터관리가 되고 있다. 사회에서 유통되는 고지서의 형태는 수만 가지 이상이며, 이를 자동화, 디지털화, 데이터관리를 위해서는 고지서 인식이 필수적이다. 현재 다양한 고지서들을 관리하기 위해서 OCR(Optical Character Recognition) 기술을 활용한다. 이때, 정확도를 높이기 위해, 먼저 고지서 양식을 인식하면, OCR 인식 시 더 높은 인식률을 가질 수 있다. 본 논문에서는 고지서 양식을 구분하기 위해 인덱스로 사용할 수 있는 로고를 객체 인식하였으며, 이때 로고의 크기가 전체 고지서 대비 작으므로 딥러닝 기술 중 FPN(Feature Pyramid Network)을 작은 객체 검출에 활용하였다. 결과적으로, 제안하는 알고리즘을 통해서 자원 낭비를 줄이고, OCR 인식 정확도를 높일 수 있었다.

• 반도체디스플레이기술학회지
• /
• 제19권2호
• /
• pp.11-15
• /
• 2020

In a typical initial setup of a PCB component inspection system, operators should manually input various information such as category, position, and inspection area for each component to be inspected, thus causing much inconvenience and longer setup time. Although there are many deep learning based object detectors, RetinaNet is regarded as one of best object detectors currently available. In this paper, a method using an extended RetinaNet is proposed that automatically detects its component category and position for each component mounted on PCBs from a high-resolution color input image. We extended the basic RetinaNet feature pyramid network by adding a feature pyramid layer having higher spatial resolution to the basic feature pyramid. It was demonstrated by experiments that the extended RetinaNet can detect successfully very small components that could be missed by the basic RetinaNet. Using the proposed method could enable automatic generation of inspection areas, thus considerably reducing the setup time of PCB component inspection systems.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권1호
• /
• pp.60-79
• /
• 2022

The U-Net architecture-based segmentation models attained remarkable performance in numerous medical image segmentation missions like skin lesion segmentation. Nevertheless, the resolution gradually decreases and the loss of spatial information increases with deeper network. The fusion of adjacent layers is not enough to make up for the lost spatial information, thus resulting in errors of segmentation boundary so as to decline the accuracy of segmentation. To tackle the issue, we propose a new deep learning-based segmentation model. In the decoding stage, the feature channels of each decoding unit are concatenated with all the feature channels of the upper coding unit. Which is done in order to ensure the segmentation effect by integrating spatial and semantic information, and promotes the robustness and generalization of our model by combining the atrous spatial pyramid pooling (ASPP) module and channel attention module (CAM). Extensive experiments on ISIC2016 and ISIC2017 common datasets proved that our model implements well and outperforms compared segmentation models for skin lesion segmentation.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2019년도 춘계학술발표대회
• /
• pp.398-400
• /
• 2019

Faced with the increasing demand for image recognition on mobile devices, how to run convolutional neural network (CNN) models on mobile devices with limited computing power and limited storage resources encourages people to study efficient model design. In recent years, many effective architectures have been proposed, such as mobilenet_v1, mobilenet_v2 and PeleeNet. However, in the process of feature selection, all these models neglect some information of shallow features, which reduces the capture of shallow feature location and semantics. In this study, we propose an effective framework based on Feature Pyramid Networks to improve the recognition accuracy of deep and shallow images while guaranteeing the recognition speed of PeleeNet structured images. Compared with PeleeNet, the accuracy of structure recognition on CIFA-10 data set increased by 4.0%.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2021년도 추계학술대회
• /
• 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.

• 방송공학회논문지
• /
• 제26권1호
• /
• pp.14-25
• /
• 2021

서로 다른 시간에 촬영된 같은 위치의 원격 탐사 영상에서 변화된 사항을 찾는 변화 탐지는 다양한 영역에 적용되기 때문에 매우 중요하다. 그러나 정합 오차, 건물 변위 오차, 그림자 오차 등이 오탐지를 발생시킨다. 이러한 문제점을 해결하기 위해 본 논문은 CADNet(Change Attention Dense Siamese Network)을 제안한다. CADNet은 다양한 크기의 변화 영역을 탐지하기 위해 FPN(Feature Pyramid Network)을 사용하며, 변화 영역에 주목하는 변화 주목 모듈을 적용하고, 낮은 수준 (Low-level)의 특징과 높은 수준 (High-level)의 특징을 모두 포함하고 있는 피처 맵을 변화 탐지에 사용하기 위해 DenseNet을 피처 추출기로 사용한다. CADNet의 성능을 Precision, Recall, F1 측면에서 측정하였을 때 WHU 데이터 세트에 대하여 98.44%, 98.47%, 98.46%이었고, LEVIR-CD 데이터 세트에 대해 90.72%, 91.89%, 91.30%이었다. 이 실험의 결과는 CADNet이 기존 변화 탐지 방법들보다 향상된 성능을 제공한다는 것을 보여준다.