A5010297685- 3D Shape Modeling and Analysis
- Epigenetics and DNA Methylation
- Single-cell and spatial transcriptomics
- Medical Imaging and Analysis
- Computer Graphics and Visualization Techniques
- Human Pose and Action Recognition
- Multimodal Machine Learning Applications
- Medical Imaging Techniques and Applications
- Disaster Management and Resilience
- Spine and Intervertebral Disc Pathology
- Topic Modeling
- Computational Geometry and Mesh Generation
- Machine Learning and Data Classification
- Data-Driven Disease Surveillance
- Anomaly Detection Techniques and Applications
- Computational and Text Analysis Methods
- Advanced Vision and Imaging
- Cell Image Analysis Techniques
- Gait Recognition and Analysis
- Hand Gesture Recognition Systems
- Domain Adaptation and Few-Shot Learning
Tsinghua University
2018-2025
University of Toronto
2023
Vector Institute
2023
University of Pittsburgh
2023
National University of Singapore
2005-2006
This work considers gradient-based mesh optimization, where we iteratively optimize for a 3D surface by representing it as the isosurface of scalar field, an increasingly common paradigm in applications including photogrammetry, generative modeling, and inverse physics. Existing implementations adapt classic extraction algorithms like Marching Cubes or Dual Contouring; these techniques were designed to extract meshes from fixed, known fields, optimization setting they lack degrees freedom...
In this paper, we investigate the problem of RGB-D egocentric action recognition. Unlike conventional human videos that are passively recorded by static cameras, self-generated from wearable sensors more flexible and provide close-ups with visual attention wearers when they act. Moreover, contain spatial appearance temporal information in RGB modality reflect 3D structure scenes depth modality. To adequately learn nonlinear heterogeneous representations different modalities exploit their...
Neural radiance fields achieve unprecedented quality for novel view synthesis, but their volumetric formulation remains expensive, requiring a huge number of samples to render high-resolution images. Volumetric encodings are essential represent fuzzy geometry such as foliage and hair, they well-suited stochastic optimization. Yet, many scenes ultimately consist largely solid surfaces which can be accurately rendered by single sample per pixel. Based on this insight, we propose neural that...
In this paper, we investigate the problem of group activity recognition by learning semantics-perserving attention and contextual interaction among different people. Conventional methods usually aggregate features extracted from individual persons pooling operations, which lack physical meaning cannot fully explore information for recognition. To address this, develop a Semantics-Preserving Teacher-Student (SPTS) networks architecture. Our SPTS first learn Teacher Network in semantic domain...
<title>Abstract</title> Large-scale foundation models have recently opened a new avenue to artificial general intelligence for life sciences, showing great promise in the analysis of single-cell transcriptomic data. Nevertheless, such challenges as tremendous number signaling regions, extreme data sparsity, and nearly binary nature epigenomic prevented construction model epigenomics thus far, though it is evident that abundant properties chromatin accessibility provide more decisive insights...
Large-scale foundation models have recently opened new avenues for artificial general intelligence. Such a research paradigm has shown considerable promise in the analysis of single-cell sequencing data, while to date, efforts centered on transcriptome. In contrast gene expression, chromatin accessibility provides more decisive insights into cell states, shaping regulatory landscapes that control transcription distinct types. Yet, challenges also persist due abundance features, high data...
Abstract BERT is a pre-trained language model that achieves state-of-the-art performance on natural processing (NLP) tasks. Once it was published, quickly became one of the most popular models in NLP field. The official recommended method for applying to downstream tasks fine-tuning. However, we argue transfer learning also very practical approach BERT, and this has its own advantages compared In paper, explore through uses fine-tuning approaches separately apply same eight GLUE benchmark...
The escalating food insecurity in Africa, caused by factors such as war, climate change, and poverty, demonstrates the critical need for advanced early warning systems. Traditional methodologies, relying on expert-curated data encompassing climate, geography, social disturbances, often fall short due to limitations, hindering comprehensive analysis potential discovery of new predictive factors. To address this, this paper introduces "HungerGist", a multi-task deep learning model utilizing...