Live camera stream: from Raspberry Pi to your local computer host.
Live object detection: with Coral EdgeTPU on host side.
High FPS: 35-45, depending mostly on network condition.
Multiple platform: ready for Ubuntu, MacOS, Windows.
Application scenario: AI-powered surveillance camera.
- Raspberry Pi, latest RPi4 is recommended.
- Latest Raspbian is recommended for RPi.
- A camera for RPi, such as RPi camera module V2.
- A local computer host with usb 3.0 port:
- Ubuntu, MacOS, Windows are supported.
- Coral EdgeTPU. USB version is recommended.
- Python virtual env is recommended:
- virtualenv (python3.7) for RPi.
- Anaconda / Conda (python3.7) for host.
- IDE such as PyCharm is recommended for host.
Clone this repository on both RPi and computer host sides:
git clone https://github.com/redlogo/RPi-Stream.git
Install libs needed for Raspberry Pi:
bash RPi-requirements.sh
Install libs required for local computer with EdgeTPU unplugged:
# Ubuntu
bash computer-hose-requirements-linux.sh
# MacOS
bash computer-host-requirements-macos.sh
# Windows
check out computer-host-requirements-windows.txt
Plug USB EdgeTPU into the host usb 3.0 port.
Edit on RPi side, change sender_stream.py:
# line 20, change it to your local computer host ip
target_ip = '192.168.7.33'
Firstly execute script on RPi side:
python3 sender_stream.py
Secondly execute script on local computer host side:
python3 receiver_stream_object_detection.py
(Windows: try 'python' w/o '3' or use IDE instead of Windows CMD)
To exit, firstly terminate RPi side, then host side.
General Python 3 Coding style.
1.0 - April 2020.
redlogo
MIT
Copyright © 2020 redlogo