Efficient Computing Backbone高效计算 Backbone
CAS-ViT and additive attention support efficient perception on embedded devices, with deployment experience across SenseTime edge terminals.以 CAS-ViT、加性注意力等为代表的高效视觉骨干,已应用于多个嵌入式端侧场景。
The center's research centers on the entire spectrum—from multimodal representation and cognitive understanding to content generation, interactive decision-making, and deployment in real-world scenarios.中心关注从多模态表征到认知理解、内容生成、交互决策与真实场景应用的完整研究链条。
CAS-ViT and additive attention support efficient perception on embedded devices, with deployment experience across SenseTime edge terminals.以 CAS-ViT、加性注意力等为代表的高效视觉骨干,已应用于多个嵌入式端侧场景。
Foundational research on cross-modal alignment, long-video reasoning, pose-language pretraining, and multimodal evaluation for international industry use cases.聚焦跨模态对齐、长视频理解、姿态语言预训练与多模态评测,形成面向海外产业应用需求的基础研究能力。
This project connects video, 3D motion, language, and interaction to study human behavior. The video presents a research thread spanning representation learning, multi-person interaction, intention analysis, long-video evaluation, and robust motion recovery.该项目连接视频、三维运动、语言与交互信息,研究复杂人类行为的理解机制。视频集中展示从表征学习、多人交互、意图分析,到长视频评测与复杂环境运动恢复的研究链条。
Project details →进入项目页面 →
Professor Huaping Zhang is the Director of the NLPIR Laboratory and Secretary-General of the Multilingual Intelligent Information Processing Technical Committee of the Chinese Association for Artificial Intelligence. His research interests include multilingual intelligent information processing, big data search and mining, natural language processing, and intelligent information analysis. 张华平教授现任NLPIR实验室主任、中国人工智能学会多语种智能信息处理专委会秘书长,主要研究方向包括多语种智能信息处理、大数据搜索与挖掘、自然语言处理以及智能信息分析。
View team → 查看团队组成 →
Professor Lei Li is Deputy Director of the Cognitive Intelligence Research Center and Secretary-General of the CAAI Multilingual Intelligent Processing Technical Committee. His research interests include large multimodal and language models, medical AI, embodied and physical intelligence, human motion understanding, and controllable generation.李磊教授现任认知智能研究中心副主任、中国人工智能学会多语种智能处理专委会秘书长,主要研究方向包括大型多模态与语言模型、医疗人工智能、具身与物理智能、人类运动理解及可控内容生成。
View team →查看团队组成 → Personal website ↗个人主页 ↗