A5030754694- Multimodal Machine Learning Applications
- Advanced Image and Video Retrieval Techniques
- Topic Modeling
- Human Pose and Action Recognition
- AI-based Problem Solving and Planning
- Emergency and Acute Care Studies
- Machine Learning and Data Classification
- Fluid Dynamics and Mixing
- Spectroscopy and Chemometric Analyses
- Heat Transfer and Boiling Studies
- Healthcare Operations and Scheduling Optimization
- Robotic Path Planning Algorithms
- Hospital Admissions and Outcomes
- Heavy Metals in Plants
- Domain Adaptation and Few-Shot Learning
- Innovative Microfluidic and Catalytic Techniques Innovation
- Image Retrieval and Classification Techniques
- Tea Polyphenols and Effects
- Advanced Graph Neural Networks
- Robot Manipulation and Learning
Shanghai Jiao Tong University
2023-2025
Stanford University
2024
Georgia Institute of Technology
2019-2024
Task-oriented grasping (TOG) refers to the problem of predicting grasps on an object that enable subsequent manipulation tasks. To model complex relationships between objects, tasks, and grasps, existing methods incorporate semantic knowledge as priors into TOG pipelines. However, is typically constructed based closed-world concept sets, restraining generalization novel concepts out pre-defined sets. address this issue, we propose GraspGPT, a large language (LLM) framework leverages open-end...
Autonomous service robots require computational frameworks that allow them to generalize knowledge new situations in a manner models uncertainty while scaling real-world problem sizes. The Robot Common Sense Embedding (RoboCSE) showcases class of frameworks, multi-relational embeddings, have not been leveraged robotics model semantic knowledge. We validate RoboCSE on realistic home environment simulator (AI2Thor) measure how well it generalizes learned about object affordances, locations,...
In this work, the bubble merger process in a heated symmetric microfluidic T-junction is numerically studied with variations of heat flux and seed volume. Detailed behaviors phase change transfer characteristics are revealed. Results show that experiences slipping colliding regimes at small large volumes, respectively. The grows faster Obvious peak evaporation rate during can be seen. significantly affect transfer. asymmetry under regime leads to difference between two main channel walls....
Background: Accurately predicting waiting time for patients is crucial effective hospital management. The present study examined the prediction of outpatient in a Chinese pediatric through use machine learning algorithms. If are informed about their advance, they can make more decisions and better plan visit on day admission. Methods: First, novel classification method clinic was proposed, which based medical knowledge statistical analysis. Subsequently, four algorithms [linear regression...
Objects rarely sit in isolation everyday human environments. If we want robots to operate and perform tasks our environments, they must understand how the objects manipulate will interact with structural elements of environment for all but simplest tasks. As such, we'd like reason about multiple environmental relate one another those relations may change as robot interacts world. We examine problem predicting inter-object object-environment between previously unseen novel environments purely...
3D visual grounding is a challenging task that often requires direct and dense supervision, notably the semantic label for each object in scene. In this paper, we instead study naturally supervised setting learns from only scene QA pairs, where prior works underperform. We propose Language-Regularized Concept Learner (LARC), which uses constraints language as regularization to significantly improve accuracy of neuro-symbolic concept learners setting. Our approach based on two core insights:...