Awesome-NCNN

NCNN is a high-performance neural network inference framework optimized for the mobile platform. This repo lists some awesome ncnn-based projects. Welcome Star & Fork & Pull Requests!

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。本仓库 收集了基于NCNN的很棒的项目。欢迎 star & fork & pull request 一键三连！

Applications 应用类

针对具体任务、场景的基于ncnn的项目

ncnn-picture-enhancement ncnn照片去雾和水下增强

A simple demo to run dehaze / underwater model in Android

https://github.com/JuZiSYJ/ncnn-picture-enhancement

Android/iOS camera preview with YOLOv5

移动端目标检测，当前项目使用的是YOLOv5的5s模型，摄像头实时捕获视频流进行检测

https://github.com/WZTENG/YOLOv5_NCNN

face mask detection on Rasberry Pi 树莓派上的口罩检测

https://github.com/Qengineering/Face-Mask-Detection-Raspberry-Pi-64-bits

YOLOv5 NCNN Implementation

YOLOv5 on Android based on NCNN

https://github.com/sunnyden/YOLOV5_NCNN_Android

DAIN ncnn Vulkan

DAIN, Depth-Aware Video Frame Interpolation implemented with ncnn library

https://github.com/nihui/dain-ncnn-vulkan

PFLD pytorch Implementation ，自带 ncnn 推理实现

Practical Facial Landmark Detector with PyTorch and NCNN implementation.

https://github.com/polarisZhao/PFLD-pytorch

C++ implemententation of LFFD with ncnn

LFFD ( A Light and Fast Face Detector for Edge Devices )'s implementation in NCNN.

https://github.com/SyGoing/LFFD-with-ncnn

基于ncnn的android端的enet分割

https://github.com/watersink/enet-as-linux

mobile-lpr: 一个面向移动端的准商业级车牌识别库

https://github.com/xiangweizeng/mobile-lpr

Iris landmarks localization 瞳孔定位，有转ncnn模型脚本

https://github.com/ItchyHiker/Iris_Landmarks_PyTorch

ultraface android 工程

https://github.com/oaup/ncnn-android-ultraface

单人人体姿态定位 android 工程

https://github.com/ZtoYtoQ/NCNN_Android_SinglePoseEstimation

Realtime Pose Estimation NCNN ONNX

https://github.com/ZtoYtoQ/NCNN-PoseEstimation

dbface ncnn 人脸检测

https://github.com/yuanluw/DBface_ncnn_demo

借鉴AlexeyAB大神的 darknet 做适量修改，用于人脸检测以及关键点检测,支持ncnn推理

https://github.com/ouyanghuiyu/darknet_face_with_landmark

ncnn在Android的一个测试,包含了人脸检测(face detection),人脸属性(face attributes),人脸识别(face recognition);车辆检测(Vehicle detection),车牌检测(plate detection),车牌识别(plate recognition);人头检测(head detection)的流程

https://github.com/791136190/ncnn_android_face_vehicle

Objects as Points ncnn 实现

https://github.com/wanglaotou/centernet_ncnn

centerface android 工程

https://github.com/JuZiSYJ/centerface-ncnn

deepsort tracking demo

https://github.com/ProLing1994/demo_deepsort

超轻量级中文ocr，支持竖排文字识别, 支持ncnn推理

https://github.com/ouyanghuiyu/chineseocr_lite

waifu2x converter ncnn version, runs fast on intel / amd / nvidia GPU with vulkan

https://github.com/nihui/waifu2x-ncnn-vulkan

ncnn style transfer android example

https://github.com/nihui/ncnn-android-styletransfer

1MB lightweight face detection model (1MB轻量级人脸检测模型)

https://github.com/Linzaer/Ultra-Light-Fast-Generic-Face-Detector-1MB

A collection of ncnn examples: face/mask detection, tracking, recognition...

https://github.com/MirrorYuChen/ncnn_example

Tools 工具类

模型转换工具、对ncnn封装等相关项目

darknet ncnn android 工程

https://github.com/paleomoon/darknet-ncnn-android

ROS wrapper for NCNN neural inference framework

https://github.com/nilseuropa/ros_ncnn

pyncnn: python wrapper of ncnn with pybind11

https://github.com/caishanli/pyncnn

Caffe INT8 Quantization convert tool

https://github.com/BUG1989/caffe-int8-convert-tools

NCNN lite without C++ support.

https://github.com/nullptr-leo/ncnn-lite

Waifu2x filter for VapourSynth

https://github.com/Nlzy/vapoursynth-waifu2x-ncnn-vulkan

SRMD super resolution for VapourSynth

https://github.com/Kiyamou/VapourSynth-SRMD-ncnn-Vulkan

ncnn .NET wrapper written in C++ and C# for Windows, MacOS and Linux

https://github.com/takuya-takeuchi/NcnnDotNet