Skip to content

Latest commit

 

History

History
64 lines (51 loc) · 2.12 KB

README_korean.md

File metadata and controls

64 lines (51 loc) · 2.12 KB

My Tensorflow Benchmark

개인적인 TensorFlow Benchmark 입니다. 
성능 측정에 사용된 모든 코드는 이 Repo에 공개되어있습니다.
질문/PR/Issue 언제든지 환영입니다.

Result Screent Shot (2019/06/24)

Tested Language

  • Python
  • C/C++

Tested CPU & GPU

  • Nvidia RTX2080 (8Gb)
  • GeForce GTX 1050 Mobile (4Gb)
  • AMD Ryzen 2700
  • Intel(R) Core(TM) i7-7700HQ CPU @ 2.80GHz
  • 128-core NVIDIA Maxwell™ architecture-based GPU
  • Quad-core ARM® A57
  • ARM Cortex-A53 1.4GHz

Tested Framework

  • TensorFlow
  • TensorFlow-Lite

Tested Archtecture + CPU/GPU + Framework + Language + Precision

  • x86_64 + TensorFlow + GPU + Python API + Float32
  • x86_64 + TensorFlow + CPU + Python API + Float32
  • x86_64 + TensorFlow + GPU + C API + Float32
  • x86_64 + TensorFlow + CPU + C API + Float32
  • x86_64 + TensorFlow-Lite + CPU + Python API + Float32
  • x86_64 + TensorFlow-Lite + CPU + Python API + Int8
  • aarch64 + TensorFlow + GPU + Python API + Float32
  • aarch64 + TensorFlow-Lite + CPU + Python API + Float32
  • aarch64 + TensorFlow-Lite + CPU + Python API + Int8
  • aarch64 + TensorFlow-Lite + CPU + C++ API + Float32
  • aarch64 + TensorFlow-Lite + CPU + C++ API + Int8
  • armv7l + TensorFlow + CPU + Python API + Float32
  • armv7l + TensorFlow-Lite + CPU + C++ API + Float32
  • armv7l + TensorFlow-Lite + CPU + C++ API + Int8

Google spreadsheet

Tools

Run

Build

You can build all benchmark tool using the source code in this repo.

Dataset

MS-COCO Validataion 2017

Models

SSD Mobilenet v2