Sunny-wong / DeepCamera

Deep Learning video processing surveillance on ARM GPU for face recognition and much more on the way. Turn existing camera into AI-powered. Production level platform for edge AI using ARM GPU/NPU, leveraging AutoML. The first of the world Edge AI full stack platform for developer/kids/home/SMB/enterprise/cloud, baking by community.

What's SharpAI

Full stack system for the deep learning edge computing devices, espeicailly set-up-box off the shell with image burning or Android apk installation.

Feature List

• Hige accurate Face Recognition
• Face Detection
• Inference on ARM Mali GPU
• Support Android TF Lite(GPU/CPU/NPU)
• Support open source embedded linux
• Control from mobile application
• Management System for devices
• Push Notification to Mobile Device
• Object Detection
• Distributed System based on celery
• Plugin to process video by Shinobi CCTV
• Application on Android to decode video with hw acc
• Motion Detection with Android GPU
• Lable and train from Mobile to Edge Device

Supported Device

• Raspberry Pi
• X86
• Rockchip RK3399 (Linux)
• Rockchip RK3288 (Android)
• Samsung 7420 (Android)
• Rockchip RK3399 (Android)
• Linux/Android Has OpenCL (TVM Supported)

Supported Camera

• Dahua Camera
• Hikvision Camera
• Shinobi CCTV Supported Devices
• Screen Captured from Android Camera preview application

How to use it

Install Prebuilt Image on tested hardware

https://github.com/SharpAI/facebox_sdk

Build From Source

coming soon

Run on Laptop

coming soon

Call For Help

• Documents, A LOT OF DOCUMENTS, we already deploy our deep camera in industry leading company but we don't have extra resource to build up community friendly documents.
• Tutorial on youtube, we can build up some of them, only if with your help, we can success together.
• Translation, we did deploy our production in China MainLand, much of our resource is in Chinese, need your help to translate, especially the Mobile APP built with Meteor https://github.com/SharpAI/mobile_app_server (need i18n selection based on system language)
• Remove unused code/project in https://github.com/SharpAI/mobile_app_server
• PR, more than welcome, anything :)
• Meetup hosts, in Silicon Valley

How to test on PC

git clone https://github.com/SharpAI/sharpai
cd sharpai/docker
docker-compose -f docker-compose-x86.yml up

Runtime On AARCH64

SharpAI On Android AARCH64

1. Download Termux Modified Version on PC

git clone https://github.com/SharpAI/mobile_app_server -b android_porting

2. Add authorized key

Copy your pc ~/.ssh/id_rsa.pub to android ~/.ssh/authorized_keys (using ssh to connect android device)

AndroidPorting/Launcher/app/src/main/assets/authorized_keys

3. Setup on Android

pkg install openssh
sshd

4. Then Remote access it for easy

ssh -p 8022 username@192.168.x.x

5. Install Base Root File System

ON PC

百度云盘

uploading

scp -P 8022 usr_aarch64_dev_1204_2018.tgz a@192.168.x.x:/data/data/com.termux/files/

On Android

cd /data/data/com.termux/files
tar -xvf usr_aarch64_dev_1204_2018.tgz

6. Run Sharp AI Code

Download Source Code

git clone https://github.com/SharpAI/sharpai
cd sharpai
./setup.sh (rk3288 运行 ./setup_arm32.sh)

Start Service

./start_service.sh

RTSP Input

Use RTSP Decoder

编译，打包

安装Pyinstaller

pip2 download pyinstaller
tar -xjvf PyInstaller-3.4.tar.bz2
cd PyInstaller-3.4
sed -i'' -e 's#"/usr/tmp"#"/data/data/com.termux/files/usr/tmp"#g' bootloader/src/pyi_utils.c
CFLAGS="-I/data/data/com.termux/files/usr/include/libandroid-support" LDFLAGS="-landroid-support" pip2 install .

Build

cd build
bash ./build_aarch64.sh runtime's path

Run

cd [runtime full path]/runtime
bash ./start_aarch64.sh

SharpAI on Android ARM32(RK3288)

Development

因为RK3288的data只有3G，需要加一个SD卡，一旦编译完毕，可以 tar 到备份SD卡，本地做删除，才有空间继续编译开发。

Termux的执行环境（开发用）

usr_dev_root_1128_2018.tgz

包含了除SVM之外的编译运行依赖，这是开发用的

链接:https://pan.baidu.com/s/1MjlCUiiUVf0z_ILoZ7y44w 密码:3rh7

空间不够了，这样节省一些：

pkg uninstall gcc-6 gcc-7 gcc-8

Use sharpai/build/build_arm.sh to Build

Arch Linux的执行环境（开发用）

arch_dev_root.tgz

ARCH Linux才能够正常使用scikit-learn(SVM),原因没查出来

链接:https://pan.baidu.com/s/1TJzKemhjfk_CWqbxaz7nvw 密码:b5cg

cd ~
tar -zxvf arch_dev_root.tgz

wget https://sdrausty.github.io/TermuxArch/setupTermuxArch.sh
bash setupTermuxArch.sh

./arch/startarch

Use sharpai/build/build_arm_svc.sh to Build

打包后的程序运行环境（产品发布用）

Termux Runtime: runtime_termux_armv7.tgz

链接:https://pan.baidu.com/s/136d1nVtPfQrrxqCZWjebLA 密码:5e53

编译后的可执行程序

链接:https://pan.baidu.com/s/1x71O1npURpMvQCv-jQ4Fwg 密码:qyex

cd ~
tar -zxvf runtime_all_armv7.tgz

Arch Linux Runtime: runtime_arch_linux_armv7.tgz

链接:https://pan.baidu.com/s/16ta4yC_mp6AOrhMyCs6N0w 密码:xwdr

cd ~
tar -zxvf runtime_arch_linux_armv7.tgz

wget https://sdrausty.github.io/TermuxArch/setupTermuxArch.sh
bash setupTermuxArch.sh

./arch/startarch c "cd /data/data/com.termux/files/home/runtime_arch/bin && ./classifier "