- Generative Adversarial Networks and Image Synthesis
- Advanced Image Processing Techniques
- Computer Graphics and Visualization Techniques
- Adversarial Robustness in Machine Learning
- Statistical Methods and Inference
- Digital Media Forensic Detection
- Machine Learning in Healthcare
- Privacy-Preserving Technologies in Data
- Reinforcement Learning in Robotics
- Machine Learning and Data Classification
- Image Retrieval and Classification Techniques
- Hydrological Forecasting Using AI
- 3D Shape Modeling and Analysis
- Privacy, Security, and Data Protection
- Coastal wetland ecosystem dynamics
- Biochemical effects in animals
- Cryptography and Data Security
- Image and Signal Denoising Methods
- Constraint Satisfaction and Optimization
- Data Quality and Management
- Video Analysis and Summarization
- Topic Modeling
- Advanced Neural Network Applications
- Grey System Theory Applications
- Image Processing Techniques and Applications
University of California, Santa Barbara
2023
University of Michigan
2018-2022
Michigan United
2020
The Affiliated Hospital to Changchun University of Chinese Medicine
2017
Northeast Agricultural University
2014
Generative models have been widely studied in computer vision. Recently, diffusion drawn substantial attention due to the high quality of their generated images. A key desired property image generative is ability disentangle different attributes, which should enable modification towards a style without changing semantic content, and parameters generalize Previous studies found that adversarial networks (GANs) are inherently endowed with such disentanglement capability, so they can perform...
Diffusion-based models have achieved state-of-the-art performance on text-to-image synthesis tasks. However, one critical limitation of these is the low fidelity generated images with respect to text description, such as missing objects, mismatched attributes, and mislocated objects. One key reason for inconsistencies inaccurate cross-attention in both spatial dimension, which controls at what pixel region an object should appear, temporal how different levels details are added through...
There are no practical and effective mechanisms to share high-dimensional data including sensitive information in various fields like health financial intelligence or socioeconomics without compromising either the utility of exposing private personal secure organizational information. Excessive scrambling encoding makes it less useful for modelling analytical processing. Insufficient preprocessing may compromise introduce a substantial risk re-identification individuals by stratification...
Health advances are contingent on continuous development of new methods and approaches to foster data-driven discovery in the biomedical clinical sciences. Open-science team-based scientific offer hope for tackling some difficult challenges associated with managing, modeling, interpreting large, complex, multisource data. Translating raw observations into useful information actionable knowledge depends effective domain-independent reproducibility, area-specific replicability, data curation,...
Diffusion-based models have achieved state-of-the-art performance on text-to-image synthesis tasks. However, one critical limitation of these is the low fidelity generated images with respect to text description, such as missing objects, mismatched attributes, and mislocated objects. One key reason for inconsistencies inaccurate cross-attention in both spatial dimension, which controls at what pixel region an object should appear, temporal how different levels details are added through...
Meta reinforcement learning (meta-RL) aims to learn a policy solving set of training tasks simultaneously and quickly adapting new tasks. It requires massive amounts data drawn from infer the common structure shared among Without heavy reward engineering, sparse rewards in long-horizon exacerbate problem sample efficiency meta-RL. Another challenge meta-RL is discrepancy difficulty level tasks, which might cause one easy task dominating thus preclude adaptation This work introduces novel...
Chinese herbal medicine (CHM) has been used for treating insomnia centuries. The most CHM was Polygonum multiflorum. However, the molecular mechanism preventing is unknown. Stilbene glucoside (THSG), an important active component of P. multiflorum, may play role insomnia. To test hypothesis, Kunming mice were treated with different dosages THSG. examine sleep duration, a computer-controlled sleep-wake detection system implemented. Electroencephalogram (EEG) and electromyogram (EMG)...
Abstract Health advances are contingent on continuous development of new methods and approaches to foster data driven discovery in the biomedical clinical health sciences. Open-science offers hope for tackling some challenges associated with Big Data team-based scientific discovery. Domain-independent reproducibility, area-specific replicability, curation, analysis, organization, management sharing health-related digital objects critical components. This study expands functionality utility...
In blurry images, the degree of image blurs may vary drastically due to different factors, such as varying speeds shaking cameras and moving objects, well defects camera lens. However, current end-to-end models failed explicitly take into account diversity blurs. This unawareness compromises specialization at each blur level, yielding sub-optimal deblurred images redundant post-processing. Therefore, how specialize one model simultaneously levels, while still ensuring coverage...
Vision language models (VLMs) are an exciting emerging class of (LMs) that have merged classic LM capabilities with those image processing systems. However, the ways these combine not always intuitive and warrant direct investigation. One understudied capability in VLMs is visual spatial planning -- ability to comprehend arrangements objects devise action plans achieve desired outcomes scenes. In our study, we introduce VSP, a benchmark 1) evaluates general, 2) breaks down task into...
The disentanglement of StyleGAN latent space has paved the way for realistic and controllable image editing, but does know anything about temporal motion, as it was only trained on static images? To study motion features in StyleGAN, this paper, we hypothesize demonstrate that a series meaningful, natural, versatile small, local movements (referred to "micromotion", such expression, head movement, aging effect) can be represented low-rank spaces extracted from conventionally pre-trained...
Meta reinforcement learning (meta-RL) aims to learn a policy solving set of training tasks simultaneously and quickly adapting new tasks. It requires massive amounts data drawn from infer the common structure shared among Without heavy reward engineering, sparse rewards in long-horizon exacerbate problem sample efficiency meta-RL. Another challenge meta-RL is discrepancy difficulty level tasks, which might cause one easy task dominating thus preclude adaptation This work introduces novel...
Generative models have been widely studied in computer vision. Recently, diffusion drawn substantial attention due to the high quality of their generated images. A key desired property image generative is ability disentangle different attributes, which should enable modification towards a style without changing semantic content, and parameters generalize Previous studies found that adversarial networks (GANs) are inherently endowed with such disentanglement capability, so they can perform...