News

微信公众号:"集成智能感知系统"

集成智能感知系统应用,算法,架构和电路系统

运行维护:清华大学集成智能视觉感知研究组

二维码:集成智能感知系统-二维码

Graduate and Intern Students Positions. Welcome on board!

iVip研究组招收本科实习生及合作研究生

iVip课题组的研究工作是设计未来的集成智能感知系统,面向机器人感知、穿戴式感知和环境信息感知等应用场景。目前的研究工作包括设计以机器人应用为目标的感知系统平台;以及设计面向未来机器人等极端性能要求下的新型高能效硬件架构和软硬件协同设计方法。

Contact: qiaofei@tsinghua.edu.cn

方向 I:智能感知系统平台

iSensing Systems: intelligent Sensing Systems

高能效智能感知系统是未来机器人的“眼睛”,“鼻子”,“耳朵”和“皮肤”,甚至“大脑”,帮助机器人和各种无人系统在多变的动态非结构环境中完成感知和情景理解,并进一步自主作出决策判断,执行自然的交互功能。目标是结合机器人应用场景,设计具有多种传感方式的智能感知系统,并探索算法精度、处理速度和能效的优化空间。

1. 面向多种机器人感知通道(视觉,听觉和触觉等)的环境感知算法设计,包括基于经典的特征提取算法,深度机器学习算法和多模态信息融合算法。

2. 机器人系统感知硬件集成和优化,包括采用FPGA和嵌入式GPU等高能效计算设备实现算法优化。

3. 针对复杂的的动态非结构环境,设计面向机器人等无人驾驶系统的智能感知应用,包括定位导航,目标识别跟踪,以及自然人机交互。

Keywords: 机器视觉, OpenCV, 立体视觉, 三维重建, 信息融合, 深度机器学习, FPGA, Verilog, GPU, 定位导航, 路径规划, SLAM, 多种传感器集成

方向 II:高能效感知计算架构和未来智能集成传感系统

Future Sensors: Perceptual Computing Architectures and Integrated Circuits

伴随着摩尔定律的终结,集成电路工艺的进步不能持续支持未来具有高计算复杂度的智能感知系统的设计和实现,需要探索新型硬件计算架构。面向图像处理和深度机器学习等具有误差容忍特性和运算概率特性的应用场景,在传统的速度,功耗和芯片面积的设计空间中引入计算质量这一维度,采用近似计算架构以取得在特定应用场景下能效的显著提升。进一步,提出物理计算架构技术尝试突破现有计算系统范式,面向多媒体智能传感系统,通过在传感信号域直接提取信息的方法,期望获得3~6个数量级的系统性能和能效提升。

1. 算法设计和评估,采用新型计算架构后对算法精度结果的影响,以及算法的优化。

2. 新型计算架构设计和实现,包括近似计算和物理计算。

Keywords: 机器视觉, 深度机器学习, 科学计算, 近似计算, 物理计算, 电路仿真, 混合信号集成电路设计, 数字集成电路设计, Matlab, C++


20170425 Congrats!

Analog-to-Information signal processing method and integrated circuits are adopted as the energy efficient implementation for Neural Network tasks.

Kaige Jia(Master student, THU), etc. "AICNN: Implementing Typical CNN Algorithms with Analog-to-Information Conversion Architecture", ISVLSI 2017.

Hong Liu(Master student, THU), etc. "AIsim: Functional Simulator for Analog-to-Information Perceptual Systems", ISVLSI 2017.


20170218 Congrats!

Qin Li(Ph.D. student, THU), etc. "From 'MISSION: IMPOSSIBLE' to Mission Possible: Fully Flexible Intelligent Contact Lens for Image Classification with Analog-to-Information Processing", ISCAS 2017.

This is a co-authored paper with Prof. Xing Wu's group, from Shanghai Key Laboratory of Multidimensional Information Processing, Dept. of Electrical Engineering, East China Normal University, Shanghai, China.


20170214 Congrats!

Yuanchang Chen(Master student, THU) and Yizhe Zhu(Intern student, BUPT), etc. "Evaluating Data Resilience in CNNs from an Approximate Memory Perspective", GLSVLSI 2017.


20161031 Congrats!

Xinghua Yang(Ph.D. candidate, THU) and Nanyang Huang(Intern student, BUPT), as joint first authors, "A priority-based selective bit dropping strategy to reduce DRAM and SRAM power in image processing", IEICE Electronics Express.


20160416 Congrats!

Make dreams come true. Computable flexible circuits and systems for coming wearable electronics. Nan Wu(Undergraduate student, THU) and Zheyu Liu(Ph.D. student, THU), as joint first authors, “A Real-Time and Energy-Efficient Implementation of Difference-of-Gaussian with Flexible Thin-Film Transistors”,ISVLSI 2016.


20160301 Congrats!

Collaboration work published. Approximate computing methods have been adopted to process remote sensing data. Yuanfeng Wu(Associate Professor, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences), Xinghua Yang(Ph.D. candidate, THU), "Approximate Computing of Remotely Sensed Data: SVM Hyperspectral Image Classification as a Case Study", IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING.


Copyright © 2016, iVip Group, NICS Lab,Dept. of EE, Tsinghua University

Beijing P.R.C All rights reserved