GitHub - Tencent/ncnn: ncnn is a high-performance neural network inference framework optimized for the mobile platform

标签: | 发表时间:2019-01-26 16:42 | 作者:
出处:https://github.com

ncnn

License Build Status Coverage Status

ncnn is a high-performance neural network inference computing framework optimized for mobile platforms. ncnn is deeply considerate about deployment and uses on mobile phones from the beginning of design. ncnn does not have third party dependencies. it is cross-platform, and runs faster than all known open source frameworks on mobile phone cpu. Developers can easily deploy deep learning algorithm models to the mobile platform by using efficient ncnn implementation, create intelligent APPs, and bring the artificial intelligence to your fingertips. ncnn is currently being used in many Tencent applications, such as QQ, Qzone, WeChat, Pitu and so on.

ncnn 是一个为手机端极致优化的高性能神经网络前向计算框架。ncnn 从设计之初深刻考虑手机端的部署和使用。无第三方依赖,跨平台,手机端 cpu 的速度快于目前所有已知的开源框架。基于 ncnn,开发者能够将深度学习算法轻松移植到手机端高效执行,开发出人工智能 APP,将 AI 带到你的指尖。ncnn 目前已在腾讯多款应用中使用,如 QQ,Qzone,微信,天天P图等。


Support most commonly used CNN network

支持大部分常用的 CNN 网络

  • Classical CNN: VGG AlexNet GoogleNet Inception ...
  • Practical CNN: ResNet DenseNet SENet FPN ...
  • Light-weight CNN: SqueezeNet MobileNetV1/V2 ShuffleNetV1/V2 MNasNet ...
  • Detection: MTCNN facedetection ...
  • Detection: VGG-SSD MobileNet-SSD SqueezeNet-SSD MobileNetV2-SSDLite ...
  • Detection: Faster-RCNN R-FCN ...
  • Detection: YOLOV2 YOLOV3 MobileNet-YOLOV3 ...
  • Segmentation: FCN PSPNet ...

HowTo

how to build ncnn libraryon Linux / Windows / Raspberry Pi3 / Android / iOS

how to use ncnn with alexnetwith detailed steps, recommended for beginners :)

ncnn 组件使用指北 alexnet附带详细步骤,新人强烈推荐 :)

ncnn low-level operation api

ncnn param and model file spec

ncnn operation param weight table

how to implement custom layer step by step


FAQ

ncnn throw error

ncnn produce wrong result


Features

  • Supports convolutional neural networks, supports multiple input and multi-branch structure, can calculate part of the branch
  • No third-party library dependencies, does not rely on BLAS / NNPACK or any other computing framework
  • Pure C ++ implementation, cross-platform, supports android, ios and so on
  • ARM NEON assembly level of careful optimization, calculation speed is extremely high
  • Sophisticated memory management and data structure design, very low memory footprint
  • Supports multi-core parallel computing acceleration, ARM big.LITTLE cpu scheduling optimization
  • The overall library size is less than 500K, and can be easily reduced to less than 300K
  • Extensible model design, supports 8bit quantization and half-precision floating point storage, can import caffe/pytorch/mxnet/onnx models
  • Support direct memory zero copy reference load network model
  • Can be registered with custom layer implementation and extended
  • Well, it is strong, not afraid of being stuffed with 卷 QvQ

功能概述

  • 支持卷积神经网络,支持多输入和多分支结构,可计算部分分支
  • 无任何第三方库依赖,不依赖 BLAS/NNPACK 等计算框架
  • 纯 C++ 实现,跨平台,支持 android ios 等
  • ARM NEON 汇编级良心优化,计算速度极快
  • 精细的内存管理和数据结构设计,内存占用极低
  • 支持多核并行计算加速,ARM big.LITTLE cpu 调度优化
  • 整体库体积小于 500K,并可轻松精简到小于 300K
  • 可扩展的模型设计,支持 8bit 量化和半精度浮点存储,可导入 caffe/pytorch/mxnet/onnx 模型
  • 支持直接内存零拷贝引用加载网络模型
  • 可注册自定义层实现并扩展
  • 恩,很强就是了,不怕被塞卷 QvQ

Example project

技术交流QQ群:637093648(已满qaq) 853969140 答案:卷卷卷卷卷


License

BSD 3 Clause

相关 [github tencent ncnn] 推荐:

GitHub - Tencent/ncnn: ncnn is a high-performance neural network inference framework optimized for the mobile platform

- -
ncnn 是一个为手机端极致优化的高性能神经网络前向计算框架. ncnn 从设计之初深刻考虑手机端的部署和使用. 无第三方依赖,跨平台,手机端 cpu 的速度快于目前所有已知的开源框架. 基于 ncnn,开发者能够将深度学习算法轻松移植到手机端高效执行,开发出人工智能 APP,将 AI 带到你的指尖.

微信Mars 组件 GitHub - Tencent/mars: Mars is a cross-platform network component developed by WeChat.

- -
Mars 是微信官方的跨平台跨业务的终端基础组件. comm:可以独立使用的公共库,包括 socket、线程、消息队列、协程等;. xlog:高可靠性高性能的运行期日志组件;. STN: 信令分发网络模块,也是 Mars 最主要的部分. sample 的使用请参考. gradle 接入我们提供了两种接入方式:.

树莓派NCNN环境搭建 | 异构 AI

- -
前言镜像已经做好了,传到百度网盘中了(请大家及时保存,不定期删除. https://pan.baidu.com/s/1fhiX86L8iL8tsLbsiVa6Wg密码: e64s. SD卡要求:至少16GB,板卡型号为树莓派3B+(其他型号未知). 本系列教程采用树莓派3B+开发板:. 1.4GHz 64位4核 ARM Cortex-A53 CPU.

Ncnn使用详解(2)——Android端 - DmrfCoder的博客 - CSDN博客

- -
本片文章基于你已经完成了 这篇文章的学习,主要介绍如何将写好的c代码应用到Android项目中. 系统:Ubuntu16.04 . 软件:Android Studio. 前期准备之 ndk安装. 在正式开始前我们需要先下载安装ndk,这里介绍一种简单高效的方式,打开Android Studio,然后依次点击File->Settings->Appearance&Behavior->System Settings->Android SDK,然后在SDK Tools下找到ndk,然后选中,点击apply就可以自动下载安装了,如图: .

Home · JohnLangford/vowpal_wabbit Wiki · GitHub

- -
There are two ways to have a fast learning algorithm: (a) start with a slow algorithm and speed it up, or (b) build an intrinsically fast learning algorithm.

GitHub - jgraph/drawio: Source to www.draw.io

- -
draw.io supports IE 11, Chrome 32+, Firefox 38+, Safari 9.1.x, 10.1.x and 11.0.x, Opera 20+, Native Android browser 5.1.x+, the default browser in the current and previous major iOS versions (e.g.

基于insightface训练mobilefacenet的相关步骤及ncnn转换流程

- -
经多位网友的共同实验,原方案部分情况下迭代次数稍微不足,导致最终识别率略有小差异,为了相对容易获得论文的最佳结果,对训练方案进行简单更新,实际训练也可根据数据acc训练是否已稳定来判断lr下降的迭代次数:. 适当增大softmax迭代次数,4万-->12万;. 增大arcface第一级lr0.1的迭代次数,8万-->12万;.

Windows下编译ncnn的android端的库 - 迷若烟雨的专栏 - CSDN博客

- -
ncnn是腾讯开源的一个为手机端极致优化的高性能神经网络前向计算框架,目前已在腾讯多款应用中使用. 由于开发者使用的是linux类似的环境,因此只提供了build.sh用来构建android和iOS的库,但好在提供了CMakelist.txt文件,我们可以借助CMake进行跨平台的交叉编译. 将以下代码存为build.bat文件,双击执行即可.

git和github简介(上)

- linyehui - 没做完,没准备好
在此贴上本人在Web标准化交流会6月25日北京站的主题分享. 在线PPT:http://jinjiang.github.com/slides/learning-git/. PPT源码:https://github.com/Jinjiang/slides/tree/gh-pages/learning-git.

Github使用指南(转)

- - CSDN博客推荐文章
来自:https://github.com/neuola/neuola-legacy/wiki/github%E4%BD%BF%E7%94%A8%E6%8C%87%E5%8D%97. 如果你只是想了解 github 的使用,请跳到 Github 简介一节. 作为程序员大军之一,想必大家有这样的经历吧.