A5101888899- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- COVID-19 diagnosis using AI
- Radiomics and Machine Learning in Medical Imaging
- Lung Cancer Diagnosis and Treatment
- Anomaly Detection Techniques and Applications
- Advanced Vision and Imaging
- Machine Learning and ELM
- Advanced Image and Video Retrieval Techniques
- 3D Shape Modeling and Analysis
- Speech and Audio Processing
- Computer Graphics and Visualization Techniques
- Remote-Sensing Image Classification
- Robotics and Sensor-Based Localization
- AI in cancer detection
- Face recognition and analysis
- Vibration Control and Rheological Fluids
- Text and Document Classification Technologies
- Aeroelasticity and Vibration Control
- Face and Expression Recognition
- Multimodal Machine Learning Applications
- Indoor and Outdoor Localization Technologies
- Sparse and Compressive Sensing Techniques
- Robotic Path Planning Algorithms
- Topic Modeling
Shanghai Jiao Tong University
2018-2025
Beijing Friendship Hospital
2025
Fujian University of Technology
2022-2025
Capital Medical University
2025
China University of Petroleum, East China
2019-2024
Changchun University of Science and Technology
2024
Jinling Institute of Technology
2024
Longhua Hospital Shanghai University of Traditional Chinese Medicine
2024
Shanghai University of Traditional Chinese Medicine
2024
Chinese Research Academy of Environmental Sciences
2023-2024
We develop a general approach to distill symbolic representations of learned deep model by introducing strong inductive biases. focus on Graph Neural Networks (GNNs). The technique works as follows: we first encourage sparse latent when train GNN in supervised setting, then apply regression components the extract explicit physical relations. find correct known equations, including force laws and Hamiltonians, can be extracted from neural network. our method non-trivial cosmology example-a...
Lip reading has witnessed unparalleled development in recent years thanks to deep learning and the availability of large-scale datasets. Despite encouraging results achieved, performance lip reading, unfortunately, remains inferior one its counterpart speech recognition, due ambiguous nature actuations that makes it challenging extract discriminant features from movement videos. In this paper, we propose a new method, termed as by Speech (LIBS), which goal is strengthen recognizers. The...
FPGA-based CNN accelerators have advantages in flexibility and power efficiency so are being deployed by a number of cloud computing service providers, including Microsoft, Amazon, Tencent, Alibaba. Given the increasing complexity neural networks, however, it is becoming challenging to efficiently map CNNs multi-FPGA platforms. In this work, we present scalable framework, FPDeep, which helps engineers specific CNN's training logic cluster or build RTL implementations for target network. With...
Lip reading aims at decoding texts from the movement of a speaker's mouth. In recent years, lip methods have made great progress for English, both word-level and sentence-level. Unlike however, Chinese Mandarin is tone-based language relies on pitches to distinguish lexical or grammatical meaning, which significantly increases ambiguity task. this paper, we propose Cascade Sequence-to-Sequence Model (CSSMCM) reading, explicitly models tones when predicting sentence. Tones are modeled based...
In recent years, researchers pay growing attention to the few-shot learning (FSL) task address data-scarce problem. A standard FSL framework is composed of two components: i) Pre-train. Employ base data generate a CNN-based feature extraction model (FEM). ii) Meta-test. Apply trained FEM novel (category different from data) acquire embeddings and recognize them. Although have made remarkable breakthroughs in FSL, there still exists fundamental Since with usually cannot adapt class...
Few-shot learning (FSL), purposing to resolve the problem of data-scarce, has attracted considerable attention in recent years. A popular FSL framework contains two phases: (i) pre-train phase employs base data train a CNN-based feature extractor. (ii) meta-test applies frozen extractor novel (novel different categories from data) and designs classifier for recognition. To correct few-shot distribution, researchers propose Semi-Supervised Few-Shot Learning (SSFSL) by introducing unlabeled...
In this work, we propose a voxel-based single-stage fine-grained and efficient point cloud 3D object detection algorithm to address the inadequate granularity in feature extraction tasks imbalance between efficiency accuracy scenarios. We develop lightweight multibranch cross-sparse convolution network (LMCCN) that is designed preserve of original while achieving enhanced efficiency. Additionally, introduce compact self-attention augmented bird's eye view (BEV) module (CFSAM). This aims...
Stereo matching is a key technique for metric depth estimation in computer vision and robotics. Real-world challenges like occlusion non-texture hinder accurate disparity from binocular cues. Recently, monocular relative has shown remarkable generalization using foundation models. Thus, to facilitate robust stereo with cues, we incorporate model into the recurrent stereo-matching framework, building new framework model-based stereo-matching, DEFOM-Stereo. In feature extraction stage,...
Extracting buildings from high-resolution remote sensing images is currently a research hotspot in the field of applications. Deep learning methods have significantly improved accuracy building extraction, but there are still deficiencies such as blurred edges, incomplete structures and loss details extraction results. To obtain accurate contours clear boundaries buildings, this article proposes novel method utilizing multi-scale attention gate enhanced positional information. By employing...
The rapid development of antibiotic resistance is occurring at a global scale. We therefore stride into the post-antibiotic era and have to battle in Anthropocene. Metals are widely used their pollution widespread worldwide. More importantly, metal-induced co-selection greatly expands environmental resistomes increases health risk environments. Here, we reviewed increasingly important roles resistance. In particular, highlight metal-rich environments that maintain reservoirs for high-risk...
Gossamer space structures technology have gained widely applications in missions. However, the vibration problem is a great challenge which makes complicated. The overall motivation of this work to develop control system for gossamer structures. In study, membrane structure with piezoelectric stack actuators bracketed on its support frame considered. First, description smart and dynamic model are presented. Then, decentralized adaptive fuzzy method developed vibration. Finally, experimental...
This paper deals with the active vibration control of smart truss structure. First, electro-mechanical coupled dynamic model structure is constructed. Then, first-order ordinary differential equation system presented. After that, an online learning fuzzy (OLFC) algorithm proposed to vibrations. The OLFC composed a reward function, Q algorithm, rule base generator and conventional controller. learns by interaction plant, changes generate policy via evaluative signal realize goal. only needs...
Terrestrial gross primary productivity (GPP) is the major carbon input to terrestrial ecosystem. The Yangtze River Basin (YRB) holds a key role in shaping China’s economic and social progress, as well ecological environmental protection. However, how GPP YRB responds climate factors remain unclear. In this research, we applied Vegetation Photosynthesis Model (VPM) data explore spatial temporal variations of during 2000–2018. Based on China Meteorological Forcing Dataset (CMFD), partial least...