行動裝置識別與追蹤物體之關鍵技術-仿神經智慧視覺系統晶片
Enabling Technology of Object Recognition and Tracking for Mobile Devices – Neuromorphic Intelligent Vision System-on-Chip
計畫團隊成員 Members 鄭桂忠教授
Prof. Kea-Tiong (Samuel) Tang
國立清華大學電機系
EE Dept., NTHU
謝志成教授
Prof. Chih-Cheng Hsieh
國立清華大學電機系
EE Dept., NTHU
羅中泉教授
Prof. Chung-Chuan Lo
國立清華大學系神所
LSSN Dept., NTHU
呂仁碩副教授
Assoc. Prof.
Ren-Shuo Liu
國立清華大學電機系
EE Dept., NTHU
Category
Processing-In-Sensor, Computing-In-Memory, Neuromorphic chip, Object tracking and prediction, Configurable-bitwidth CNN, Run-time bitwidth flexibility, Power-accuracy tradeoff
技術亮點
Technical Highlights
本計畫專注於智慧行動裝置(如智慧手機、無人機和機器人)的智能技術。更具體地說,我們使用以下仿生和神經形態算法、架構和電路技術,來設計並實現於低電壓、低功耗和高速限制下執行目標檢測、避障、深度估計等的晶片與系統。
- 智慧感測器運算(CIS)晶片
- 記憶體內運算(CIM)晶片
- 突波神經網路(SNN) 演算法和晶片
- 受果蠅啟發的 AI 視覺模組和演算法
- 融合數位與類比之AI晶片架構
This project focuses on the enabling techniques of intelligent mobile devices such as smartphones, drones, and robots. More specifically, we design and implement chips and systems that perform objection detection, obstacle avoidance, depth estimation, etc. under low-voltage, low-power, and high-speed constraints using the following bio-inspired and neuromorphic algorithms, architecture, and circuits techniques.
- Intelligent compute-in-sensor (CIS) chips
- Intelligent compute-in-memory (CIM) chips
- Spiking neural network (SNN) algorithms and chips
- Fruit fly-inspired AI vision modules and algorithms
- Analog-digital-integrated AI chips and architecture