ICML 2026 [26.6% acceptance rate]
代表论文
WACV 2026 [33.7% acceptance rate]
VividAnimator: An End-to-End Audio and Pose-driven Half-Body Human Animation Framework
AAAI 2026 [17.6% acceptance rate]
FashionMAC: Deformation-Free Fashion Image Generation with Fine-Grained Model Appearance Customization
IEEE T-PAMI 2025 [IF 20.8]
Physically Based Facial Texture Generation In The Wild
NeurIPS 2025 [24.52% acceptance rate]
UniTransfer: Video Concept Transfer via Progressive Spatio-Temporal Decomposition
SIGGRAPH 2025 [22.6% acceptance rate]
BuildingBlock: A Hybrid Approach for Structured Building Generation
SCIENCE CHINA Information Sciences 2025 [IF 7.6]
Artificial intelligence for virtual reality: a review
CVPR 2025 [21.12% acceptance rate]
AnimateAnything: Consistent and Controllable Animation for video generation
AAAI 2022 (Oral) [Oral presentation 5.5% acceptance rate; overall 15.0%]
Active Boundary Loss for Semantic Segmentation
ACM MM 2021 [27.9% acceptance rate]
Attention-guided Temporally Coherent Video Object Matting
所有论文
CG-MLLM: Captioning and Generating 3D Content via Multi-modal Large Language Models
ICML 2026 [26.6% acceptance rate]
VividAnimator: An End-to-End Audio and Pose-driven Half-Body Human Animation Framework
WACV 2026 [33.7% acceptance rate]
FashionMAC: Deformation-Free Fashion Image Generation with Fine-Grained Model Appearance Customization
AAAI 2026 [17.6% acceptance rate]
Physically Based Facial Texture Generation In The Wild
IEEE T-PAMI 2025
UniTransfer: Video Concept Transfer via Progressive Spatio-Temporal Decomposition
NeurIPS 2025 [24.52% acceptance rate]
研究之路
工作经历
- — 至今
研究方向: 人工智能内容生成、计算机图形学、三维视觉、以及数字孪生 - —
实习内容：
- 探索AIGC在场景生成中的应用,主要着眼于与过程建模(Procedural Modeling)方法的结合
- —
实习内容：
- 探索AIGC，主要着眼于三维生成。
- 相关技术总结产出论文一篇(PAMI第一次Revision中)，国家发明专利一项
- —
- —
实习内容:
- 探索和开发高质量的抠图和语义分割算法
- 使用公司账户在知名大规模语义分割数据集ADE20K中取得第一名
- 相关技术总结产出AAAI(Oral)一篇，MM一篇，国家发明专利一项
公告
2023 · 字体设计与生成
阿里妈妈刀隶体
阿里妈妈刀隶体，通过机器学习了爨宝子碑的精髓，强调了三角点折弯与横画两端“雁尾”之方强，外露锋芒，耀其精神，介于隶楷之间，保持着浓重的隶书方笔意味，给人以坚实、朴茂、稚拙、倔犟之感，打破常规常态的匀称均衡，结字极为大胆，大小常常出人意料，看似失度，实则深得艺理，险中求正，静中寓动，动中求静。
2023 · 字体设计与生成
阿里妈妈东方大楷体
阿里妈妈东方大楷是基于颜真卿的多宝塔碑算法生成并结合人力完成的一套字库，机器学习到了颜体的丰腴厚重之感，在笔画的描摹上做到了传统书法的收放起伏，规范了骨架、笔画，还原了书法中的落笔、转笔、藏锋、护尾等书写韵味，使整体更饱满、圆融、浑厚。