检索结果-内蒙古大学图书馆

Deep Learning Fast MRI Using Channel Attention in Magnitude Domain

学校读者我要写书评

暂无评论

Deep Learning Fast MRI Using Channel Attention in Magnitude ...

IEEE International Symposium on Biomedical Imaging

作者： Joonhyung Lee Hyunjong Kim HyungJin Chung Jong Chul Ye Department of Bio and Brain Engineering KAIST Korea Robotics Program KAIST Korea KAIST Korea

ISBN: (数字)9781538693308

ISBN: (纸本)9781538693315

Magnetic resonance imaging (MRI) acquisition is an inherently slow process whose acceleration has been the subject of much investigation. In recent years, the explosive advance of deep learning techniques for computer vision and image reconstruction has led to the investigation of deep neural networks for the reconstruction of MRI with under-sampled k-space. In this work, we propose a new image domain architecture that directly produces a sum-of-squares image from under-sampled multi-coil MRI acquisition. This model, called BarbellNet, is a fully convolutional neural network architecture that utilizes the channel attention mechanism using the residual channel attention block (RCAB). Through extensive experiments with the fastMRI data set, we confirm the efficacy of BarbellNet.

关键词： Training Magnetic resonance imaging Image reconstruction Machine learning Measurement Acceleration Feature extraction

Balanced Depth Completion between Dense Depth Inference and Sparse Range Measurements via KISS-GP

学校读者我要写书评

暂无评论

Balanced Depth Completion between Dense Depth Inference and ...

2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

作者： Sungho Yoon Ayoung Kim Robotics Program KAIST Daejeon S. Korea KAIST Daejeon S. Korea

ISBN: (数字)9781728162126

ISBN: (纸本)9781728162133

Estimating a dense and accurate depth map is the key requirement for autonomous driving and robotics. Recent advances in deep learning have allowed depth estimation in full resolution from a single image. Despite this impressive result, many deep-learning-based monocular depth estimation (MDE) algorithms have failed to keep their accuracy yielding a meter-level estimation error. In many robotics applications, accurate but sparse measurements are readily available from Light Detection and Ranging (LiDAR). Although they are highly accurate, the sparsity limits full resolution depth map reconstruction. Targeting the problem of dense and accurate depth map recovery, this paper introduces the fusion of these two modalities as a depth completion (DC) problem by dividing the role of depth inference and depth regression. Utilizing the state-of-the-art MDE and our Gaussian process (GP) based depth-regression method, we propose a general solution that can flexibly work with various MDE modules by enhancing its depth with sparse range measurements. To overcome the major limitation of GP, we adopt Kernel Interpolation for Scalable Structured (KISS)-GP and mitigate the computational complexity from O(N 3 ) to O(N). Our experiments demonstrate that the accuracy and robustness of our method outperform state-of-the-art unsupervised methods for sparse and biased measurements.

关键词： Visualization Interpolation Image resolution Simultaneous localization and mapping Laser radar Robustness Kernel

Balanced depth completion between dense depth inference and sparse range measurements via KISS-GP

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Yoon, Sungho Kim, Ayoung Robotics Program KAIST Daejeon Korea Republic of Department of Civil and Environmental Engineering KAIST Daejeon Korea Republic of

Estimating a dense and accurate depth map is the key requirement for autonomous driving and robotics. Recent advances in deep learning have allowed depth estimation in full resolution from a single image. Despite this impressive result, many deep-learning-based monocular depth estimation (MDE) algorithms have failed to keep their accuracy yielding a meter-level estimation error. In many robotics applications, accurate but sparse measurements are readily available from Light Detection and Ranging (LiDAR). Although they are highly accurate, the sparsity limits full resolution depth map reconstruction. Targeting the problem of dense and accurate depth map recovery, this paper introduces the fusion of these two modalities as a depth completion (DC) problem by dividing the role of depth inference and depth regression. Utilizing the state-of-the-art MDE and our Gaussian process (GP) based depth-regression method, we propose a general solution that can flexibly work with various MDE modules by enhancing its depth with sparse range measurements. To overcome the major limitation of GP, we adopt Kernel Interpolation for Scalable Structured (KISS)-GP and mitigate the computational complexity from O(N3) to O(N). Our experiments demonstrate that the accuracy and robustness of our method outperform state-of-the-art unsupervised methods for sparse and biased measurements. Copyright © 2020, The Authors. All rights reserved.

关键词： Optical radar

Game-theoretic model predictive control with data-driven identification of vehicle model for head-to-head autonomous racing

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Jung, Chanyoung Lee, Seungwook Seong, Hyunki Finazzi, Andrea Shim, David Hyunchul School of Electrical Engineering Korea Advanced Institute of Science and Technology Daejeon Korea Republic of KAIST Robotics Program Korea Advanced Institute of Science and Technology Daejeon Korea Republic of

Resolving edge-cases in autonomous driving, head-to-head autonomous racing is getting a lot of attention from the industry and academia. In this study, we propose a game-theoretic model predictive control (MPC) approach for head-to-head autonomous racing and data-driven model identification method. For the practical estimation of nonlinear model parameters, we adopted the hyperband algorithm, which is used for neural model training in machine learning. The proposed controller comprises three modules: 1) game-based opponents’ trajectory predictor, 2) high-level race strategy planner, and 3) MPC-based low-level controller. The game-based predictor was designed to predict the future trajectories of competitors. Based on the prediction results, the high-level race strategy planner plans several behaviors to respond to various race circumstances. Finally, the MPC-based controller computes the optimal control commands to follow the trajectories. The proposed approach was validated under various racing circumstances in an official simulator of the Indy Autonomous Challenge. The experimental results show that the proposed method can effectively overtake competitors, while driving through the track as quickly as possible without collisions. © 2021, CC BY-NC-SA.

关键词： Model predictive control

PrimA6D: Rotational Primitive Reconstruction for Enhanced and Robust 6D Pose Estimation

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Jeon, Myung-Hwan Kim, Ayoung Department of the Robotics Program KAIST Daejeon Korea Republic of Department of Civil and Environmental Engineering KAIST Daejeon Korea Republic of

— In this paper, we introduce a rotational primitive prediction based 6D object pose estimation using a single image as an input. We solve for the 6D object pose of a known object relative to the camera using a single image with occlusion. Many recent state-of-the-art (SOTA) two-step approaches have exploited image keypoints extraction followed by PnP regression for pose estimation. Instead of relying on bounding box or keypoints on the object, we propose to learn orientation-induced primitive so as to achieve the pose estimation accuracy regardless of the object size. We leverage a Variational AutoEncoder (VAE) to learn this underlying primitive and its associated keypoints. The keypoints inferred from the reconstructed primitive image are then used to regress the rotation using PnP. Lastly, we compute the translation in a separate localization module to complete the entire 6D pose estimation. When evaluated over public datasets, the proposed method yields a notable improvement over the LINEMOD, the Occlusion LINEMOD, and the YCB-Video dataset. We further provide a synthetic-only trained case presenting comparable performance to the existing methods which require real images in the training phase. Copyright © 2020, The Authors. All rights reserved.

关键词： Machine learning

SPriorSeg: Fast Road-Object Segmentation using Deep Semantic Prior for Sparse 3D Point Clouds

学校读者我要写书评

暂无评论

SPriorSeg: Fast Road-Object Segmentation using Deep Semantic...

IEEE International Conference on Systems, Man and Cybernetics

作者： Ki-In Na Byungjae Park Jong-Hwan Kim Intelligent Robotics Research Division ETRI Daejeon Republic of Korea Robotics Program KAIST Daejeon Republic of Korea School of Mechanical Engineering KOREATECH Cheonan Chungnam Republic of Korea School of Electrical Engineering KAIST Daejeon Republic of Korea

ISBN: (数字)9781728185262

ISBN: (纸本)9781728185279

Detection and classification of road-objects like cars, pedestrians, and cyclists is the first step in autonomous driving. In particular, point-wise object segmentation for 3D point clouds is essential to estimate the precise appearances of the road-objects. In this paper, we propose SPriorSeg, a fast and accurate point-level object segmentation for point clouds by integrating the strengths of deep convolutional auto-encoder and region growing algorithm. Semantic segmentation using the light-weighted convolutional auto-encoder generates semantic prior by labeling a spherical projection image of point clouds pixel-by-pixel with classes of road-objects. The region growing algorithm achieves pixel-wise instance segmentation by taking into account semantic prior and geometric features between neighboring pixels. We build a well-balanced, pixel-level labeled dataset for all classes using 3D bounding boxes and point clouds from the KITTI object dataset. The dataset is employed to train our light-weighted neural network for semantic segmentation and demonstrate the performance of both semantic and instance segmentation of SPriorSeg.

关键词： Image segmentation Three-dimensional displays Laser radar Semantics Object segmentation Labeling Autonomous vehicles

Three-Dimensional Visual Mapping of Underwater Ship Hull Surface using View-based Piecewise-Planar Measurements

学校读者我要写书评

暂无评论

IFAC-PapersOnLine 2019年第21期52卷 384-389页

作者： Seonghun Hong Jinwhan Kim Robotics Program KAIST Daejeon Korea Department of Mechanical Engineering KAIST Daejeon Korea

Visual inspection of underwater ship hulls using unmanned underwater vehicles needs to be conducted at very close range to the target surface because of the limited visibility conditions in underwater environments due to light attenuation, scattering, and turbidity. These environmental challenges result in ineffective photometric and geometric information in hull surface images, and thus, the performance of conventional 3D reconstruction techniques is often not satisfactory. This paper presents a visual mapping method for 3D reconstruction of underwater ship hull surface using a monocular camera as a primary mapping sensor. The main idea of the proposed approach is to model the moderately curved hull surface as a combination of piecewise-planar panels, and to generate a global map by aligning the images in a 2D reference frame and correcting them appropriately to reflect the information of perspective projections of the 3D panels. Experimental results are shown to demonstrate the feasibility of the proposed method using a dataset obtained in a field experiment with a full-scale ship in a real sea environment.

关键词： visual mapping 3D reconstruction computer vision autonomous underwater vehicle visual inspection

Time-to-Line Crossing Enhanced End-to-End Autonomous Driving Framework

学校读者我要写书评

暂无评论

Time-to-Line Crossing Enhanced End-to-End Autonomous Driving...

International Conference on Intelligent Transportation

作者： Chanyoung Jung Hyunki Seong David Hyunchul Shim School of Electrical Engineering Korea Advanced Institute of Science and Technology Daejeon South Korea KAIST Robotics Program Korea Advanced Institute of Science and Technology Daejeon South Korea

ISBN: (数字)9781728141497

ISBN: (纸本)9781728141503

End-to-end autonomous driving approach, which directly maps raw input images to vehicle control commands using a deep neural network, is gaining considerable attention from both academia and industry. Researchers have conducted studies on this subject over the past few years. However, they have focused on designing network architecture, and evaluated performance only with root mean square error (RMSE), which did not account for the temporal dynamics of autonomous driving. In this paper, we propose a time-to-line crossing (TLC) enhanced end-to-end driving framework. The proposed framework is original to the industry in three ways. First, for the fine-labeled training dataset for end-to-end autonomous driving, the TLC based label correction algorithm is applied to reduce the inaccuracies of driver action, which is used as a ground truth. Second, we designed a novel deep neural network model based on bi-directional convolutional long short-term memory (bi-CLSTM) which can sufficiently encode the spatial and temporal features in the input image sequence. Third, in addition to the RMSE evaluation metric, we validated the performance of the end-to-end driving model using TLC, the advanced driving support system (ADAS), and a driver performance indicator. We integrated the proposed framework with existing end-to-end driving models on a full-scale autonomous vehicle in the experimental portion of our study. The results show that the proposed framework is valid and that our network model outperforms the existing models in terms of both RMSE and TLC.

关键词： Feature extraction Autonomous vehicles Training Vehicles Network architecture Data mining Wheels

Balancing domain gap for object instance detection

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Yun, Woo-han Lee, Jaeyeon Kim, Jaehong Kim, Junmo Robotics Program KAIST and Intelligent Robotics Research Division ETRI Korea Republic of Intelligent Robotics Research Division ETRI Korea Republic of School of Electrical Engineering KAIST Korea Republic of

Object instance detection in cluttered indoor environment is a core functionality for service robots. We can readily build a detection system by following recent successful strategy of deep convolutional neural networks, if we have a large annotated dataset. However, it is hard to prepare such a huge dataset in instance detection problem where only small number of samples are available. This is one of main impediment to deploying an object detection system. To overcome this obstacle, many approaches to generate synthetic dataset have been proposed. These approaches confront the domain gap or reality gap problem stems from discrepancy between source domain (synthetic training dataset) and target domain (real test dataset). In this paper, we propose a simple approach to generate a synthetic dataset with minimum human effort. Especially, we identify that domain gaps of foreground and background are unbalanced and propose methods to balance these gaps. In the experiment, we verify that our methods help domain gaps to balance and improve the accuracy of object instance detection in cluttered indoor environment. Copyright © 2019, The Authors. All rights reserved.

关键词： Deep neural networks