DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 유회준 | - |
dc.contributor.author | Kwon, Beomseok | - |
dc.contributor.author | 권범석 | - |
dc.date.accessioned | 2024-07-30T19:31:37Z | - |
dc.date.available | 2024-07-30T19:31:37Z | - |
dc.date.issued | 2024 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1097214&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/321642 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전기및전자공학부, 2024.2,[iii, 17 p. :] | - |
dc.description.abstract | Human pose estimation (HPE) is a promising solution for accurately understanding the state and context of human actions in virtual reality (VR). A high frame rate with low-power HPE processing is required for a realistic user interaction experience in battery-limited mobile devices. The proposed HPE accelerator is a computing-in-memory (CIM) based accelerator that computes depth-wise separable convolution (DWSC) of a lightweight HPE network. Three key features contribute to a resource-efficient CIM accelerator: 1) Dual-mode CIM computes DWSC with a reconfigurable homogenous architecture, resulting in $2.68$ times higher throughput than previous analog CIMs. 2) Effective layer-aware unrolling performs bit-parallel computation on dual-mode CIM with fewer ADC operations, achieving 46 times higher throughput than before. 3) Adaptive fused intermacro balancing improves latency balance in layer fusion execution, leading to a $57.0 %$ higher frame rate than before. The proposed HPE accelerator is implemented in $28nm$ CMOS technology. It achieves higher computation resource utilization and operates HPE with a low energy-delay product of $27.6 uJ \cdot s$ in mobile VR devices. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | 메모리 내 컴퓨팅▼a인간 포즈 추정▼a축전기 기반 아날로그 연산▼a재구성 가능한 동종 아키텍처▼aSRAM | - |
dc.subject | computing-in-memory(CIM)▼ahuman pose estimation▼acapacitor-based analog computation▼areconfigurable homogeneous architecture▼aSRAM | - |
dc.title | (A) computing-in-memory-based human pose estimation accelerator with resource-efficient macro for mobile devices | - |
dc.title.alternative | 모바일 디바이스를 위한 리소스 효율적인 메모리 내 연산 기반 인간 포즈 추정 가속기 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전기및전자공학부, | - |
dc.contributor.alternativeauthor | Yoo, Hoi-jun | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.