Home Original page

GitHub - u4ea2/DeepCamera: AI Face Recognition/Person Detection NVR. Machine Learning On The Edge, turn your Camera into AI-powered with Jetson Nano and telegram to protect your privacy.

DeepCamera is a sample application from sharpai-hub.

Get started on X86 Machine(Windows/Linux/Mac)

Installation

  • Register account on DeepCamera website
  • Install sharpai-hub: pip3 install sharpai-hub
  • Login on device: sharpai-cli login
  • Register device: sharpai-cli device register
  • Start DeepCamera: sharpai-cli deepcamera start

Connect To Camera through NVR Gui

Tested Camera:

  • DaHua / Lorex / AMCREST: URL Path: /cam/realmonitor?channel=1&subtype=0 Port: 554
  • Ip Camera Lite on IOS: URL Path: /live Port: 8554

If you are using a camera but have no idea about the RTSP URL, please file an issue or use iSpyConnect to get camera streaming URL format

Features

  • FFMpeg with Nvidia Nano hardware decoder
  • Face Detector with Nvidia Nano GPU TensorRT MTCNN
  • Face Embedding with Nvidia Nano GPU Pytorch InsightFace
  • Person Detection with GPU
  • Integrate with telegram bot API
  • Porting to Jetson Nano
  • High 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
  • Native raspberry pi camera support
  • Labelling server and application is down, need BYOD document API server repo
  • Image upload to AWS or on premise AWS compatiable server(MINIO)

Todos

  • Integration with Home Assistant
  • Simplify installation script

Commercial Version

  • Provide real time pipeline on edge device
  • E2E pipeline to support model customization
  • Cluster on the edge
  • Port to specific edge device/chipset
  • Voice application (ASR/KWS) end to end pipeline
  • ReID model
  • Behavior analysis model
  • Transformer model
  • Contrastive learning
  • Click to join sharpai slack channel for commercial support

DeepCamera Architecture

architecture

Demo On Youtube

Demo On Youtube

Get Started on Jetson Nano

Install Docker-compose

sudo apt-get install -y libhdf5-dev python3 python3-pip
pip3 install -U pip
sudo pip3 install docker-compose==1.27.4

Get source code

git clone https://github.com/SharpAI/DeepCamera

Create Token for Telegram Bot

  • Create Telegram Bot through @BotFather
  • Set Telegram Token in Configure File
  • Send message to the new bot you created

Start DeepCamera

cd DeepCamera  
./run-on-nano.sh start

Connect To Camera through RTSP URL

On Jetson Nano, Access to 8080 port. http://localhost:8080
Default username and password is:
username: user@sharpaibox.com
password: SharpAI2018 Tested Camera: DaHua / Lorex / AMCREST, URL Path: /cam/realmonitor?channel=1&subtype=0 Port: 554 You can use iSpyConnect to get camera streaming URL When setup done, you will see live view on web page, when detected person in camera, you will receive video clips on telegram.

Label on Web GUI, train face recognition model on device

cat docker/workaipython/ro_serialno 
82f28703d001

82f28703d001 is device ID.
Access http://165.232.62.29:3000/ how to config on web gui

Detail information

Camera streaming URL format

Use Mobile APP to label and train face recognition model on device

Get device serial number

cat docker/workaipython/ro_serialno 
82f28703d001

82f28703d001 is device ID.
Generate QRCode of device ID

If your have any question or feature request, please feel free to join slack for commercial support

Slack

Click to join sharpai slack channel

Contributions