1,037 research outputs found
생성 모델링을 사용한 컨볼루션 신경망을 이용한 카테터 위치 결정 및 추적
No full textCatheter Segmentation, Catheter Synthesis, Domain Adaptation, deep convolutional neural network, Semantic Segmentation1. Motivation, Research Problem and Contributions 1
1 Motivation 1
2 Main Contributions 2
3 Thesis Outline 4
4 Publications 5
2. Catheter Tip Tracking in Camera Sequences 7
1 Introduction 7
2 Related Work 9
3 Detect and Segment to Track 11
3.1 Detection Network 12
3.2 Segmentation Network 14
4 Experiments and Results 14
4.1 Dataset 14
4.2 Evaluation Settings 15
4.3 Quantitative Results 17
4.4 Qualitative Results 20
5 Discussion 21
5.1 Effectiveness of the proposed methods 21
5.2 Effect of data augmentation 21
6 Chapter Summary 22
3. Camera Catheter Translation to X-ray Catheter Sequences 23
1 Introduction 23
2 Related Work 25
2.1 Learning based methods for segmentation and detection 26
2.2 Image translation in medical imaging 26
3 Synthesize and Segment 27
3.1 Synthesize: GAN based X-ray translation 27
3.2 Segment: From synthesis to segmentation 29
4 Experiments 30
4.1 Datasets 30
4.2 Experimental Setup 31
4.3 Quantitative Results 32
4.4 Qualitative Results 34
5 Discussion 38
6 Chapter Summary 39
4. Domain Adaptive Segmentation: Generated X-ray Catheter to Real X-ray Catheter 41
1 Introduction 41
2 Related Work 43
2.1 Video Semantic Segmentation 43
2.2 Domain Adaptation for Semantic Segmentation 44
3 Learning Domain Adaptation for Semantic Segmentation 45
3.1 Problem Definition 45
3.2 Overview 45
3.3 Video Segmentation with PCM 47
4 Catheter Dataset Results 49
4.1 Catheter Dataset Experimental Setting 49
4.2 Comparative Analysis on Catheter Dataset 51
5 Cityscapes Dataset Results 52
5.1 Cityscapes Experimental Setting 52
5.2 Comparison with State-of-the-Art Methods 53
6 Discussion 57
6.1 PCM Mask Generation 57
6.2 Inconsistent Labels 58
7 Chapter Summary 58
5. Concluding Remarks and Future Work 60
6. Acknowledgement 62
References 63DoctordCollectio
Real-time Tracking of Guidewire Robot Tips using Deep Convolutional Neural Networks on Successive Localized Frames
No full textStudies are proceeded to stabilize cardiac surgery using thin micro-guidewires and catheter robots. To control the robot to a desired position and pose, it is necessary to accurately track the robot tip in real time but tracking and accurately delineating the thin and small tip is challenging. To address this problem, a novel image analysis-based tracking method using deep convolutional neural networks (CNN) has been proposed in this paper. The proposed tracker consists of two parts; (1) a detection network for rough detection of the tip position and (2) a segmentation network for accurate tip delineation near the tip position. To learn a robust real-time tracker, we extract small image patches, including the tip in successive frames and then learn the informative spatial and motion features for the segmentation network. During inference, the tip bounding box is first estimated in the initial frame via the detection network, thereafter tip delineation is consecutively performed through the segmentation network in the following frames. The proposed method enables accurate delineation of the tip in real time and automatically restarts tracking via the detection network when tracking fails in challenging frames. Experimental results show that the proposed method achieves better tracking accuracy than existing methods, with a considerable real-time speed of 19ms.1
EmoP3D: A brain like pyramidal deep neural network for emotion recognition
No full textThe paper reports a new model based on the understanding and encompassing intelligence from brain i.e. biological pyramidal neurons, tailored for emotion recognition. Our objective is to introduce and utilize usage of non-Convolutional layers in models and show comparable or state-of-the-art performance for multi-class emotion recognition problem. We open-sourced the optimized code for researchers. Our model shows state-of-the-art performance on two emotion recognition datasets (eNTERFACE and Youtube) enhancing previous best result by 9.47%9.47% and 20.8 .8%, respectively
Finite sample econometrics / Aman Ullah.
No full texteconomic&political bookfair2015Includes bibliographical references (p. 199-225) and index.x, 230 pages
Machine Learning-Enabled Power Scheduling in IoT-Based Smart Cities
No full textRecent advancements in hardware and communication technologies have enabled worldwide interconnection using the internet of things (IoT). The IoT is the backbone of smart city applications such as smart grids and green energy management. In smart cities, the IoT devices are used for linking power, price, energy, and demand information for smart homes and home energy management (HEM) in the smart grids. In complex smart grid-connected systems, power scheduling and secure dispatch of information are the main research challenge. These challenges can be resolved through various machine learning techniques and data analytics. In this paper, we have proposed a particle swarm optimization based machine learning algorithm known as a collaborative execute-before-after dependency-based requirement, for the smart grid. The proposed collaborative execute-before-after dependency-based requirement algorithm works in two phases, analysis and assessment of the requirements of end-users and power distribution companies. In the first phases, a fixed load is adjusted over a period of 24 h, and in the second phase, a randomly produced population load for 90 days is evaluated using particle swarm optimization. The simulation results demonstrate that the proposed algorithm performed better in terms of percentage cost reduction, peak to average ratio, and power variance mean ratio than particle swarm optimization and inclined block rate. © 2021 Tech Science Press. All rights reserved.1
Disturbance observer based sliding mode control for unmanned helicopter hovering operations in presence of external disturbances
No full textNumerous control techniques are developed for miniature unmanned helicopters to do hover operation with each method having its own advantages and limitations. During the hover operation helicopters suffer from unknown external disturbances such as wind and ground effect. For a stable operation, these disturbances must be compensated accurately. This paper presents a disturbance observer based sliding mode control technique for small-scale unmanned helicopters to do hover operation in presence of external disturbances. To counteract both matched and mismatched uncertainties a new sliding surface is designed based on the disturbances estimations. The controller design is based on the linearized state-space model of the helicopter which effectively describes helicopter dynamics during the hover operations. The model mismatch and external disturbances are estimated as lumped disturbances and are compensated in the controller design. The proposed controller reduces chattering and is capable of handling matched and mismatched uncertainties. The control performance is successfully tested in Simulink
- …
