- Multimodal Machine Learning Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Interactive and Immersive Displays
- Hand Gesture Recognition Systems
- Receptor Mechanisms and Signaling
- Gait Recognition and Analysis
- Tactile and Sensory Interactions
- Cell Image Analysis Techniques
- Computational Drug Discovery Methods
- Image Processing Techniques and Applications
- Optical measurement and interference techniques
- Machine Learning in Bioinformatics
- Chinese history and philosophy
- China's Ethnic Minorities and Relations
- Educational Reforms and Innovations
- Emotional Intelligence and Performance
- Modular Robots and Swarm Intelligence
- Welding Techniques and Residual Stresses
- 3D Shape Modeling and Analysis
- 3D Surveying and Cultural Heritage
- Medical Research and Treatments
- Optical Polarization and Ellipsometry
South China Normal University
2023
Columbia University
2021-2023
Universiti Sains Malaysia
2023
Guizhou University
2022
Guangdong University of Technology
2021
Indiana University – Purdue University Indianapolis
2015-2018
Central South University
2018
University of Indianapolis
2018
University Ucinf
2015
Hunan Normal University
2009-2011
Few-shot object detection (FSOD) aims to detect objects using only a few examples. How adapt state-of-the-art detectors the few-shot domain remains challenging. Object proposal is key ingredient in modern detectors. However, quality of proposals generated for classes existing methods far worse than that many-shot classes, e.g., missing boxes due misclassification or inaccurate spatial locations with respect true objects. To address noisy problem, we propose novel meta-learning based FSOD...
Few-shot object detection (FSOD) aims to detect never-seen objects using few examples. This field sees recent improvement owing the meta-learning techniques by learning how match between query image and few-shot class examples, such that learned model can generalize novel classes. However, currently, most of meta-learning-based methods perform parwise matching regions (usually proposals) classes separately, therefore failing take into account multiple relationships among them. In this paper,...
Sighted players gain spatial awareness within video games through sight and tools (SATs) such as minimaps. Visually impaired (VIPs), however, must often rely heavily on SATs to awareness, especially in complex environments where using rich ambient sound design alone may be insufficient. Researchers have developed many for facilitating VIPs. Yet this abundance disguises a gap our understanding about how exactly these approaches assist VIPs gaining what their relative merits limitations are....
We study a bio-detection application as case to demonstrate that Kmeans -- based unsupervised feature learning can be simple yet effective alternative deep techniques for small data sets with limited intra-as well inter-class diversity. investigate the effect on classifier performance of augmentation extraction multiple patch sizes and at different image scales. Our set includes 1833 images from four classes bacteria, each bacterial culture captured three wavelengths overall collected during...
We study a bio-detection application as case to demonstrate that Kmeans -- based unsupervised feature learning can be simple yet effective alternative deep techniques for small data sets with limited intra-as well inter-class diversity. investigate the effect on classifier performance of augmentation extraction multiple patch sizes and at different image scales. Our set includes 1833 images from four classes bacteria, each bacterial culture captured three wavelengths overall collected during...
In a small sample welding test space, and to achieve online prediction self-optimisation of process parameters for the resistance joint quality power lithium battery packs, this paper proposes model. The model combines chaos game optimisation algorithm (CGO) with multi-output least-squares support vector regression machine (MLSSVR), multi-objective parameter method based on particle swarm algorithm. First, MLSSVR was constructed, hyperparameter strategy CGO designed. Next, predicted using...
Selecting the right drugs for patients is a primary goal of precision medicine. In this article, we consider problem cancer drug selection in learning-to-rank framework. We have formulated as to accurately predicting 1) ranking positions sensitive and 2) orders among cell lines based on their responses drugs. developed new method, denoted pLETORg, that predicts structures each line via using latent vectors vectors. The pLETORg method learns such through explicitly enforcing that, list line,...
Few-shot object detection (FSOD) aims to detect objects using only a few examples. How adapt state-of-the-art detectors the few-shot domain remains challenging. Object proposal is key ingredient in modern detectors. However, quality of proposals generated for classes existing methods far worse than that many-shot classes, e.g., missing boxes due misclassification or inaccurate spatial locations with respect true objects. To address noisy problem, we propose novel meta-learning based FSOD...
Few-shot object detection (FSOD) aims to detect never-seen objects using few examples. This field sees recent improvement owing the meta-learning techniques by learning how match between query image and few-shot class examples, such that learned model can generalize novel classes. However, currently, most of meta-learning-based methods perform pairwise matching regions (usually proposals) classes separately, therefore failing take into account multiple relationships among them. In this...
Video games created for visually impaired players (VIPs) remain inequivalent to those sighted players. Sighted use minimaps within learn how their surrounding environment is laid out, but there no effective analogue the minimap A major accessibility challenge create a generalized, acoustic (non-visual) version of VIPs. To address this challenge, we develop and investigate four techniques which represent breadth ideas an might work: companion smartphone app, echolocation, directional scanner,...
The joint optimization of Neural Radiance Fields (NeRF) and camera trajectories has been widely applied in SLAM tasks due to its superior dense mapping quality consistency. NeRF-based learns poses using constraints by implicit map representation. A observed phenomenon that results from the this form is jerky physically unrealistic estimated motion, which turn affects quality. To address deficiency current SLAM, we propose paper TS-SLAM (TS for Trajectory Smoothness). It introduces smoothness...
Neural implicit surface reconstruction has achieved remarkable progress recently. Despite resorting to complex radiance modeling, state-of-the-art methods still struggle with textureless and specular surfaces. Different from RGB images, polarization images can provide direct constraints on the azimuth angles of normals. In this paper, we present PISR, a novel method that utilizes geometrically accurate polarimetric loss refine shape independently appearance. addition, PISR smooths normals in...
Visual commonsense understanding requires Vision Language (VL) models to not only understand image and text but also cross-reference in-between fully integrate achieve comprehension of the visual scene described. Recently, various approaches have been developed achieved high performance on benchmarks. However, it is unclear whether really underlying knowledge due limited evaluation data resources. To provide an in-depth analysis, we present a Multimodal Evaluation (ME) pipeline automatically...
Selecting the right drugs for patients is a primary goal of precision medicine. In this manuscript, we consider problem cancer drug selection in learning-to-rank framework. We have formulated as to accurately predicting 1). ranking positions sensitive and 2). orders among cell lines based on their responses drugs. developed new method, denoted pLETORg , that predicts structures each line via using latent vectors vectors. The method learns such through explicitly enforcing that, list line,...
Vision-language (VL) understanding tasks evaluate models' comprehension of complex visual scenes through multiple-choice questions. However, we have identified two dataset biases that models can exploit as shortcuts to resolve various VL correctly without proper understanding. The first type bias is \emph{Unbalanced Matching} bias, where the correct answer overlaps question and image more than incorrect answers. second \emph{Distractor Similarity} answers are overly dissimilar but...
Vision-language (VL) understanding tasks evaluate models’ comprehension of complex visual scenes through multiple-choice questions. However, we have identified two dataset biases that models can exploit as shortcuts to resolve various VL correctly without proper understanding. The first type bias is Unbalanced Matching bias, where the correct answer overlaps question and image more than incorrect answers. second Distractor Similarity answers are overly dissimilar but significantly similar...
Fromm’s ethics can be viewed as a form of virtue ethics, which emphasizes the importance developing one’s true self and reducing one's false self. By doing so, one change their psychological state in terms character tendencies, leading to an improved moral life higher standards. While some criticisms have been raised about theory, including its perceived inability guide on how act properly, Fromm's is unique that it does not focus establishing universal rules for specific actions. Instead,...
Liu Shao's Renwu Zhi is a classic work on the study of human talent in ancient China. Written during tumultuous period end Eastern Han dynasty, Shao recognized that political problems time were rooted issues related to talent. Consequently, one primary purposes his writing was assist feudal dynasties selecting and training talented individuals address crisis. This paper explores definition characteristics Caixing (talent nature) Zhi, which he accomplishes through methods Jiuzheng (nine...