A5100446112- Multimodal Machine Learning Applications
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- 3D Shape Modeling and Analysis
- Traumatic Brain Injury and Neurovascular Disturbances
- Generative Adversarial Networks and Image Synthesis
- Video Analysis and Summarization
- Computer Graphics and Visualization Techniques
- Face recognition and analysis
- Intracranial Aneurysms: Treatment and Complications
- Image Retrieval and Classification Techniques
- Facial Nerve Paralysis Treatment and Research
- Topic Modeling
- Traumatic Brain Injury Research
- Geotechnical Engineering and Analysis
- Advanced Chemical Sensor Technologies
- Optimization and Search Problems
- Domain Adaptation and Few-Shot Learning
- Construction Project Management and Performance
- Cryospheric studies and observations
- Advanced Neural Network Applications
- Nutritional Studies and Diet
- Brain Tumor Detection and Classification
- Soil Geostatistics and Mapping
- Time Series Analysis and Forecasting
First Affiliated Hospital of Jiangxi Medical College
2025
Nanchang University
2025
Nanyang Technological University
2020-2025
Tianjin Metallurgical Vocational Technical College
2024
Chinese Academy of Sciences
2015-2024
Institute of Geographic Sciences and Natural Resources Research
2021-2024
Zhejiang Gongshang University
2024
Hong Kong University of Science and Technology
2024
University of Hong Kong
2024
Xidian University
2019-2024
Food computing is playing an increasingly important role in human daily life, and has found tremendous applications guiding behavior towards smart food consumption healthy lifestyle. An task under the food-computing umbrella retrieval, which particularly helpful for health related applications, where we are interested retrieving information about (e.g., ingredients, nutrition, etc.). In this paper, investigate open research of cross-modal retrieval between cooking recipes images, propose a...
Point clouds are useful in many applications like autonomous driving and robotics as they provide natural 3D information of the surrounding environments. While there extensive research on point clouds, scene understanding 4D a series consecutive frames, is an emerging topic yet under-investigated. With (3D cloud videos), robotic systems could enhance their robustness by leveraging temporal from previous frames. However, existing semantic segmentation methods suffer low precision due to...
An important aspect of health monitoring is effective logging food consumption. This can help management diet-related diseases like obesity, diabetes, and even cardiovascular diseases. Moreover, fitness enthusiasts, people who wanting to achieve a target weight. However, food-logging cumbersome, requires not only taking additional effort note down the item consumed regularly, but also sufficient knowledge (which difficult due availability wide variety cuisines). With increasing reliance on...
Actor and action video segmentation from natural language query aims to selectively segment the actor its in a based on an input textual description. Previous works mostly focus learning simple correlation between two heterogeneous features of vision via dynamic convolution or fully convolutional classification. However, they ignore linguistic variation have difficulty modeling global visual context, which leads unsatisfactory performance. To address these issues, we propose asymmetric...
Food retrieval is an important task to perform analysis of food-related information, where we are interested in retrieving relevant information about the queried food item such as ingredients, cooking instructions, etc. In this paper, investigate cross-modal between images and recipes. The goal learn embedding recipes a common feature space, that corresponding image-recipe embeddings lie close one another. Two major challenges addressing problem 1) large intra-variance small inter-variance...
This paper investigates an open research task of text-to-image synthesis for automatically generating or manipulating images from text descriptions. Prevailing methods mainly take the textual descriptions as conditional input GAN generation, and need to train different models text-guided image generation manipulation tasks. In this paper, we propose a novel unified framework Cycle-consistent Inverse (CI-GAN) both Specifically, first model without input, aiming generate with high diversity...
Actor and action video segmentation with language queries aims to segment out the expression referred objects in video. This process requires comprehensive reasoning fine-grained understanding. Previous methods mainly leverage dynamic convolutional networks match visual semantic representations. However, convolution neglects spatial context when processing each region frame is thus challenging similar complex scenarios. To address such limitation, we construct a modulated network....
In this paper, we investigate an open research task of generating controllable 3D textured shapes from the given textual descriptions. Previous works either require ground truth caption labeling or extensive optimization time. To resolve these issues, present a novel framework, TAPS3D, to train text-guided shape generator with pseudo captions. Specifically, based on rendered 2D images, retrieve relevant words CLIP vocabulary and construct captions using templates. Our constructed provide...
Balance and gait impairments play a key role in falls among the elderly. Traditional clinical scales such as Berg Scale (BBS) to assess fall risk are often subjective, time consuming, does not performance. Shorter assessments Timed Up Go (TUG) available, but most clinicians only look into completion time. This study aimed develop fast, low-cost, automated framework for balance function assessment comprehensive analysis by enhancing traditional TUG test with markerless motion capture (MoCap)...
Currently, the authentication of virgin walnut oil (VWO) has become very important due to possible adulteration VWO with cheaper plant oils such as soybean (SO), puer tea seed (PO), and sunflower (SFO). Methods involving Fourier transform infrared (FT-IR) spectroscopy combined chemometric techniques (partial least square) were developed for quantification SO, PO, SFO in VWO. IR spectra samples recorded at frequency regions 4000–650 cm −1 on horizontal attenuated total reflectance (HATR)...
Wet weather-related hazards such as hydroplaning can be reduced with the proper use of permeable friction course (PFC). At low rainfall intensities, PFC provides quick drainage water and better skid resistance. However, at higher rates, entire volume runoff cannot discharged within porous layer, causing to occur on pavement surface. Water flow road surface result in tires. The objective study is evaluate risk multi-lane roadways using a fluid–structure interaction model. A comprehensive...
Food is significant to human daily life.In this paper, we are interested in learning structural representations for lengthy recipes, that can benefit the recipe generation and food cross-modal retrieval tasks.Different from common vision-language data, here images contain mixed ingredients target recipes paragraphs, where do not have annotations on structure information.To address above limitations, propose a novel method unsupervisedly learn sentence-level tree structures cooking...
Recently, Micro expression~(ME) has achieved remarkable progress in a wide range of applications, since it's an involuntary facial expression that reflects personal psychological state truly. In the procedure ME analysis, spotting is essential step, and non trivial to be detected from long interval video because short duration low intensity issues. To alleviate this problem, paper, we propose novel Micro- Macro-Expression~(MaE) Spotting framework based on Apex Boundary Perception...
Recently, the semantics of scene text has been proven to be essential in fine-grained image classification. However, existing methods mainly exploit literal meaning for recognition, which might irrelevant when it is not significantly related objects/scenes. We propose an end-to-end trainable network that mines implicit contextual knowledge behind and enhance correlation fine-tune representation. Unlike methods, our model integrates three modalities: visual feature extraction, correlating...
Abstract Currently, research and applications in the field of capacity prediction mainly focus on use recycling batteries, encompassing topics such as SOH estimation, RUL prediction, echelon use. However, there is scant application based battery manufacturing process. Measuring grading process an important step production. The traditional acquisition method consumes considerable time energy. To address above issues, this study establishes improved extreme learning machine (ELM) model for...