Abstract

Many recent medical segmentation systems rely on powerful deep learning models to solve highly specific tasks. To maximize performance, it is standard practice to evaluate numerous pipelines with varying model topologies, optimization parameters, pre- & postprocessing steps, and even model cascades. It is often not clear how the resulting pipeline transfers to different tasks. We propose a simple and thoroughly evaluated deep learning framework for segmentation of arbitrary medical image volumes. The system requires no task-specific information, no human interaction and is based on a fixed model topology and a fixed hyperparameter set, eliminating the process of model selection and its inherent tendency to cause method-level over-fitting. The system is available in open source and does not require deep learning expertise to use. Without task-specific modifications, the system performed better than or similar to highly specialized deep learning methods across 3 separate segmentation tasks. In addition, it ranked 5-th and 6-th in the first and second round of the 2018 Medical Segmentation Decathlon comprising another 10 tasks. The system relies on multi-planar data augmentation which facilitates the application of a single 2D architecture based on the familiar U-Net. Multi-planar training combines the parameter efficiency of a 2D fully convolutional neural network with a systematic train- and test-time augmentation scheme, which allows the 2D model to learn a representation of the 3D image volume that fosters generalization.

Abstract

Developing high performance embedded vision applications requires balancing run-time performance with energy constraints. Given the mix of hardware accelerators that exist for embedded computer vision (e.g. multi-core CPUs, GPUs, and FPGAs), and their associated vendor optimized vision libraries, it becomes a challenge for developers to navigate this fragmented solution space. To aid with determining which embedded platform is most suitable for their application, we conduct a comprehensive benchmark of the run-time performance and energy efficiency of a wide range of vision kernels. We discuss rationales for why a given underlying hardware architecture innately performs well or poorly based on the characteristics of a range of vision kernel categories. Specifically, our study is performed for three commonly used HW accelerators for embedded vision applications: ARM57 CPU, Jetson TX2 GPU and ZCU102 FPGA, using their vendor optimized vision libraries: OpenCV, VisionWorks and xfOpenCV. Our results show that the GPU achieves an energy/frame reduction ratio of 1.1-3.2x compared to the others for simple kernels. While for more complicated kernels and complete vision pipelines, the FPGA outperforms the others with energy/frame reduction ratios of 1.2-22.3x. It is also observed that the FPGA performs increasingly better as a vision application's pipeline complexity grows. Comment: 8 pages, Design Automation Conference (DAC), The 15th IEEE International Conference on Embedded Software and Systems, 2019

Abstract

Low light image enhancement is an important challenge for the development of robust computer vision algorithms. The machine learning approaches to this have been either unsupervised, supervised based on paired dataset or supervised based on unpaired dataset. This paper presents a novel deep learning pipeline that can learn from both paired and unpaired datasets. Convolution Neural Networks (CNNs) that are optimized to minimize standard loss, and Generative Adversarial Networks (GANs) that are optimized to minimize the adversarial loss are used to achieve different steps of the low light image enhancement process. Cycle consistency loss and a patched discriminator are utilized to further improve the performance. The paper also analyses the functionality and the performance of different components, hidden layers, and the entire pipeline.

Abstract

Rain removal aims to remove the rain streaks on rain images. The state-of-the-art methods are mostly based on Convolutional Neural Network~(CNN). However, as CNN is not equivariant to object rotation, these methods are unsuitable for dealing with the tilted rain streaks. To tackle this problem, we propose Deep Symmetry Enhanced Network~(DSEN) that is able to explicitly extract the rotation equivariant features from rain images. In addition, we design a self-refining mechanism to remove the accumulated rain streaks in a coarse-to-fine manner. This mechanism reuses DSEN with a novel information link which passes the gradient flow to the higher stages. Extensive experiments on both synthetic and real-world rain images show that our self-refining DSEN yields the top performance. Comment: Accepted by ICIP 19. Corresponding author: Hanrong Ye

Abstract

Recovering a high dynamic range (HDR) image from a single low dynamic range (LDR) input image is challenging due to missing details in under-/over-exposed regions caused by quantization and saturation of camera sensors. In contrast to existing learning-based methods, our core idea is to incorporate the domain knowledge of the LDR image formation pipeline into our model. We model the HDRto-LDR image formation pipeline as the (1) dynamic range clipping, (2) non-linear mapping from a camera response function, and (3) quantization. We then propose to learn three specialized CNNs to reverse these steps. By decomposing the problem into specific sub-tasks, we impose effective physical constraints to facilitate the training of individual sub-networks. Finally, we jointly fine-tune the entire model end-to-end to reduce error accumulation. With extensive quantitative and qualitative experiments on diverse image datasets, we demonstrate that the proposed method performs favorably against state-of-the-art single-image HDR reconstruction algorithms. Comment: CVPR 2020. Project page: https://www.cmlab.csie.ntu.edu.tw/~yulunliu/SingleHDR Code: https://github.com/alex04072000/SingleHDR

Abstract

Road-users are a critical part of decision-making for both self-driving cars and driver assistance systems. Some road-users, however, are more important for decision-making than others because of their respective intentions, ego vehicle's intention and their effects on each other. In this paper, we propose a novel architecture for road-user importance estimation which takes advantage of the local and global context of the scene. For local context, the model exploits the appearance of the road users (which captures orientation, intention, etc.) and their location relative to ego-vehicle. The global context in our model is defined based on the feature map of the convolutional layer of the module which predicts the future path of the ego-vehicle and contains rich global information of the scene (e.g., infrastructure, road lanes, etc.), as well as the ego vehicle's intention information. Moreover, this paper introduces a new data set of real-world driving, concentrated around inter-sections and includes annotations of important road users. Systematic evaluations of our proposed method against several baselines show promising results. Comment: Published in: IEEE Intelligent Vehicles (IV), 2019

Abstract

This paper studies the statistical nature of TV channels streaming variable bit rate distribution and allocation. The goal of the paper is to derive the best-fit rate distribution to describe TV streaming bandwidth allocation, which can reveal traffic demands of users. Our analysis uses multiplexers channel bandwidth allocation (PID) data of 13 TV live channels. We apply 17 continuous and 3 discrete distributions to determine the best-fit distribution function for each individual channel and for the whole set of channels. We found that the generalized extreme distribution fitting most of our channels most precisely according to the Bayesian information criterion. By the same criterion tlocationscale distribution matches best for the whole system. We use these results to propose parameters for streaming server queuing model. Results are useful for streaming servers scheduling policy design process targeting to improve limited infrastructural resources, traffic engineering through dynamic routing at CDN, SDN.

Abstract

utonomous driving requires 3D perception of vehicles and other objects in the in environment. Much of the current methods support 2D vehicle detection. This paper proposes a flexible pipeline to adopt any 2D detection network and fuse it with a 3D point cloud to generate 3D information with minimum changes of the 2D detection networks. To identify the 3D box, an effective model fitting algorithm is developed based on generalised car models and score maps. A two-stage convolutional neural network (CNN) is proposed to refine the detected 3D box. This pipeline is tested on the KITTI dataset using two different 2D detection networks. The 3D detection results based on these two networks are similar, demonstrating the flexibility of the proposed pipeline. The results rank second among the 3D detection algorithms, indicating its competencies in 3D detection. Comment: Accepted at ICRA 2018

Abstract

urate lane localization and lane change detection are crucial in advanced driver assistance systems and autonomous driving systems for safer and more efficient trajectory planning. Conventional localization devices such as Global Positioning System only provide road-level resolution for car navigation, which is incompetent to assist in lane-level decision making. The state of art technique for lane localization is to use Light Detection and Ranging sensors to correct the global localization error and achieve centimeter-level accuracy, but the real-time implementation and popularization for LiDAR is still limited by its computational burden and current cost. As a cost-effective alternative, vision-based lane change detection has been highly regarded for affordable autonomous vehicles to support lane-level localization. A deep learning-based computer vision system is developed to detect the lane change behavior using the images captured by a front-view camera mounted on the vehicle and data from the inertial measurement unit for highway driving. Testing results on real-world driving data have shown that the proposed method is robust with real-time working ability and could achieve around 87% lane change detection accuracy. Compared to the average human reaction to visual stimuli, the proposed computer vision system works 9 times faster, which makes it capable of helping make life-saving decisions in time.