A5004127475- Retinal Imaging and Analysis
- Genomic variations and chromosomal abnormalities
- Chromosomal and Genetic Variations
- Domain Adaptation and Few-Shot Learning
- Digital Imaging for Blood Diseases
- Retinal and Optic Conditions
- Genomics and Chromatin Dynamics
- Multimodal Machine Learning Applications
- Cell Image Analysis Techniques
- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Image Retrieval and Classification Techniques
- Colorectal Cancer Screening and Detection
- Image and Signal Denoising Methods
- Image Processing Techniques and Applications
- AI in cancer detection
- Human Pose and Action Recognition
- Brain Tumor Detection and Classification
- Stock Market Forecasting Methods
- Glaucoma and retinal disorders
- Hand Gesture Recognition Systems
- Reinforcement Learning in Robotics
- Adversarial Robustness in Machine Learning
- Interactive and Immersive Displays
- Advanced Neural Network Applications
Xi’an Jiaotong-Liverpool University
2018-2024
Suzhou University of Science and Technology
2024
University of Science and Technology of China
2024
University of Liverpool
2023
With the rapid development of robotic and AI technology in recent years, human–robot interaction has made great advancement, making practical social impact. Verbal commands are one most direct frequently used means for interaction. Currently, such can enable robots to execute pre-defined tasks based on simple explicit language instructions, e.g., certain keywords must be detected. However, that is not natural way human communicate. In this paper, we propose a novel task-based framework robot...
Traditional training methods such as card teaching, assistive technologies (e.g., augmented reality/virtual reality games and smartphone apps), DVDs, human-computer interactions, human-robot interactions are widely applied in autistic rehabilitation recent years. In this article, we propose a novel framework for human-computer/robot interaction introduce preliminary intervention study improving the emotion recognition of Chinese children with an autism spectrum disorder. The core is Facial...
Self-supervised learning is well known for its remarkable performance in representation and various downstream computer vision tasks. Recently, Positive-pair-Only Contrastive Learning (POCL) has achieved reliable without the need to construct positive-negative training sets. It reduces memory requirements by lessening dependency on batch size. The POCL method typically uses a single objective function extract distortion invariant (DIR) which describes proximity of positive-pair...
Karyotyping plays a crucial role in genetic disorder diagnosis. Currently requires considerable manual efforts, domain expertise and experience, is very time consuming. Automating the karyotyping process has been an important popular task. This study focuses on classification of chromosomes into 23 types, step towards fully automatic karyotyping. proposes convolutional neural network (CNN) based deep learning to automatically classify chromosomes. The proposed method was trained tested...
With the proposal of Segment Anything Model (SAM), fine-tuning SAM for medical image segmentation (MIS) has become popular. However, due to large size model and significant domain gap between natural images, fine-tuning-based strategies are costly with potential risk instability, feature damage catastrophic forgetting. Furthermore, some methods transferring a domain-specific MIS through disable model's prompting capability, severely limiting its utilization scenarios. In this paper, we...
Accurate fovea localization is essential for analyzing retinal diseases to prevent irreversible vision loss. While current deep learning-based methods outperform traditional ones, they still face challenges such as the lack of local anatomical landmarks around fovea, inability robustly handle diseased images, and variations in image conditions. In this paper, we propose a novel transformer-based architecture called DualStreamFoveaNet (DSFN) multi-cue fusion. This explicitly incorporates...
Foundation models have become a cornerstone in deep learning, with techniques like Low-Rank Adaptation (LoRA) offering efficient fine-tuning of large models. Similarly, methods such as Retrieval-Augmented Generation (RAG), which leverage vectorized databases, further improved model performance by grounding outputs external information. While these approaches demonstrated notable success, they often require extensive training or labeled data, can limit their adaptability resource-constrained...
Precise segmentation of chromosome in the real image achieved by a microscope is significant for karyotype analysis. The usually pixel-level classification task, which considers different instances as classes. Many instance methods predict Intersection over Union (IoU) through head branch to correct confidence. Their effectiveness based on correlation between tasks. However, none these consider input and output Herein, we propose network regression correction. First, adopt two branches...
Chromosomes carry the genetic information of humans. They exhibit non-rigid and non-articulated nature with varying degrees curvature. Chromosome straightening is an important step for subsequent karyotype construction, pathological diagnosis cytogenetic map development. However, robust chromosome remains challenging, due to unavailability training images, distorted details shapes after straightening, as well poor generalization capability. In this paper, we propose a novel architecture,...
This research looks into the applications of image captioning in Chinese. In order to improve abilities children with Autism Spectrum Disorder (ASD) spontaneous language and turn-taking, during interacting them, rehabilitation robots are used track their attention describe attractive objects scenes. method has three advantages. First, may attract ASD by describing that interest them. Second, ability which is beneficial enhancing superior expressions social communication or human beings....
The fovea is an important anatomical landmark of the retina. Detecting location essential for analysis many retinal diseases. However, robust localization remains a challenging problem, as region often appears fuzzy, and retina diseases may further obscure its appearance. This paper proposes novel Vision Transformer (ViT) approach that integrates information both inside outside to achieve localization. Our proposed network, named Bilateral-Vision-Transformer (Bilateral-ViT), consists two...
In medical imaging, chromosome straightening plays a significant role in the pathological study of chromosomes and development cytogenetic maps. Whereas different approaches exist for task, typically geometric algorithms are used whose outputs characterized by jagged edges or fragments with discontinued banding patterns. To address flaws algorithms, we propose novel framework based on image-to-image translation to learn pertinent mapping dependence synthesizing straightened uninterrupted...
The social interaction is one of the necessary skills for robots to better integrate into human society. However, current interact mainly through audio and visual means with little reliance on haptic interaction. There still exist many obstacles touch: 1) complex manufacturing process tactile sensor array main obstacle lowering cost production; 2) mode diverse. are no robot standards data sets interactive behavior in public domain. In view this, our research looks following aspects...
Invariant-based Contrastive Learning (ICL) methods have achieved impressive performance across various domains. However, the absence of latent space representation for distortion (augmentation)-related information in makes ICL sub-optimal regarding training efficiency and robustness downstream tasks. Recent studies suggest that introducing equivariance into (CL) can improve overall performance. In this paper, we rethink roles augmentation strategies improving CL efficacy. We propose a novel...
Motion control algorithms for quadruped robots undergo rapid development in recent years. Interactive have demonstrated they may positively enhance the effect of psychotherapy treatment patients with cognitive impairment, which requires them to more interactive capabilities than traditional robots. In this study, we focus on enabling imitate real animal motions extracted from videos, by design robotic motion controllers can be simplified and bionic degree enhanced. The capture data, however,...
Gaze-following is an effective way for intention understanding in human–robot interaction, which aims to follow the gaze of humans estimate what object being observed. Most existing methods require people and objects appear same image. Due limitation view camera, these are not applicable practice. To address this problem, we propose a method following that utilizes geometric map better estimation. With help map, competitive cross-frame On basis method, novel gaze-based image caption system,...
The fovea is an important anatomical landmark of the retina. Detecting location essential for analysis many retinal diseases. However, robust localization remains a challenging problem, as region often appears fuzzy, and retina diseases may further obscure its appearance. This paper proposes novel Vision Transformer (ViT) approach that integrates information both inside outside to achieve localization. Our proposed network, named Bilateral-Vision-Transformer (Bilateral-ViT), consists two...
Chromosomes carry the genetic information of humans. They exhibit non-rigid and non-articulated nature with varying degrees curvature. Chromosome straightening is an important step for subsequent karyotype construction, pathological diagnosis cytogenetic map development. However, robust chromosome remains challenging, due to unavailability training images, distorted details shapes after straightening, as well poor generalization capability. In this paper, we propose a novel architecture,...