A5026092790- Advanced Vision and Imaging
- Generative Adversarial Networks and Image Synthesis
- Advanced Image and Video Retrieval Techniques
- Brain Tumor Detection and Classification
- Machine Learning and Algorithms
- Advanced Image Processing Techniques
- Brain Metastases and Treatment
- Mobile Crowdsensing and Crowdsourcing
- Advanced Neural Network Applications
- Image Enhancement Techniques
- Image and Video Stabilization
- Advanced Bandit Algorithms Research
- Radiomics and Machine Learning in Medical Imaging
- Human Pose and Action Recognition
- Video Analysis and Summarization
- 3D Shape Modeling and Analysis
- Retinal Imaging and Analysis
- Visual Attention and Saliency Detection
- Domain Adaptation and Few-Shot Learning
- Industrial Vision Systems and Defect Detection
- Multimodal Machine Learning Applications
- Image and Video Quality Assessment
- Glioma Diagnosis and Treatment
- Cerebrovascular and Carotid Artery Diseases
- Face Recognition and Perception
Tsinghua University
2015-2022
Reichman University
2019
FACE Foundation
2019
Victoria University of Wellington
2019
This paper presents a survey of image synthesis and editing with Generative Adversarial Networks (GANs). GANs consist two deep networks, generator discriminator, which are trained in competitive way. Due to the power networks training manner, capable producing reasonable realistic images, have shown great capability many applications. surveys recent GAN papers regarding topics including, but not limited to, texture synthesis, inpainting, image-to-image translation, editing.
Background Approximately one‐fourth of all cancer metastases are found in the brain. MRI is primary technique for detection brain metastasis, planning radiotherapy, and monitoring treatment response. Progress tumor now requires new or growing at small subcentimeter size, when these therapies most effective. Purpose To develop a deep‐learning‐based approach finding metastasis on MRI. Study Type Retrospective. Sequence Axial postcontrast 3D T 1 ‐weighted imaging. Field Strength 1.5T 3T....
This paper presents a data structure that reduces approximate nearest neighbor query times for image patches in large datasets. Previous work texture synthesis has demonstrated real-time from small exemplar textures. However, high performance proved elusive modern patch-based optimization techniques which frequently use many images the tens of megapixels or above. Our new algorithm, PatchTable, offloads as much computation possible to pre-computation stage takes modest time, so patch queries...
Extracting background features for estimating the camera path is a key step in many video editing and enhancement applications. Existing approaches often fail on highly dynamic videos that are shot by moving cameras contain severe foreground occlusion. Based existing theories, we present new, practical method can reliably identify complex video, leading to accurate estimation layering. Our approach contains local motion analysis global optimization step. We first divide input into...
General image completion and extrapolation methods often fail on portrait images where parts of the human body need to be recovered - a task that requires accurate structure appearance synthesis. We present two-stage deep learning framework for tacking this problem. In first stage, given with an incomplete body, we extract complete, coherent through parsing network, which focuses recovery inside unknown region help pose estimation. second use network fill region, guided by map in stage. For...
Image matting is widely studied for accurate foreground extraction. Most algorithms, including deep-learning based solutions, require a carefully edited trimap. Recent works attempt to combine the segmentation stage and in one CNN model, but errors occurring at lead unsatisfactory matte. We propose user-guided approach practical human matting. More precisely, we provide good automatic initial natural way of interaction that reduces workload drawing trimaps allows users guide ambiguous...
Personal videos often contain visual distractors, which are objects that accidentally captured and can distract viewers from focusing on the main subjects. We propose a method to automatically detect localize these distractors through learning manually labeled dataset. To achieve spatially temporally coherent detection, we extracting features at temporal-superpixel level using traditional supporting vector machine based framework. also experiment with end-to-end convolutional neural...
Abstract We propose a novel end-to-end deep learning framework, the Joint Matting Network (JMNet), to automatically generate alpha mattes for human images. utilize intrinsic structures of body as seen in images by introducing pose estimation module, which can provide both global structural guidance and local attention focus matting task. Our network model includes network, trimap shared encoder extract features above three networks. also append refinement module gradient loss sharper matte....
We study the problem of utilizing human intelligence to categorize a large number objects. In this problem, given category hierarchy and set objects, we can ask humans check whether an object belongs category, our goal is find most cost-effective strategy locate appropriate in for each object, such that cost (i.e., questions humans) minimized. There are many important applications including image classification product categorization. develop online framework, which distribution gradually...
Human pose transfer has been becoming one of the emerging research topics in recent years. However, state-of-the-art results are still far from satisfactory. One main reason is that these end-to-end methods often blindly trained without semantic understanding its content. In this paper, we propose a novel method for human with consideration part-based representation human. particular, to segment body into multiple parts, and each them represents region With proposed layer generators,...
Abstract BACKGROUND AND OBJECTIVE: Brain metastases have been found to account for one-fourth of all cancer seen in clinics. Magnetic resonance imaging (MRI) is widely used detecting brain metastases. Accurate detection the critical design radiotherapy treat and monitor their progression or response therapy prognosis. However, finding on MRI very challenging as many are small manifest objects weak contrast images. In this work we present a deep learning approach integrated with...