bee_track

This explains how to setup a headless pi, how to setup aravis, this project and make it so the project runs on boot with no interaction

Pi setup and connection

This section explains how to setup a blank sd card and Pi4 as a headless Pi you can SSH into

• Flash OS using Raspbian Imager. You can specify the name and password of your wifi hotspot before flashing, so that the Pi will be able to access the network.
• Unplug and replug into PC
• Go to boot partition (bootfs) (NOT root/boots), to create files, by

cd /media/USERNAME/bootfs

• Create a blank file ssh.txt, by

touch ssh.txt

• Create a file wpa_supplicant.conf by

sudo nano wpa_supplicant.conf

Write the following:

ctrl_interface=DIR=/var/run/wpa_supplicant GROUP=netdev
update_config=1
country=GB

network={
	ssid="WiFi SSID"
	psk="WPA/WPA2 passphrase"
	key_mgmt=WPA-PSK
}

• Insert SD card into Pi and power on, wait until Pi turns up as connected on hotspot

• Find IP address of Pi :

ping raspberrypi.local

• SSH to Pi:

ssh pi@IP

Project setup

• First create a virtual python environment where all python and pip commands must be run:

python3 -m venv bee-venv

• Activate the virtual environment:

source bee-venv/bin/activate

• Ensure 'apt' utility is up-to-date:

sudo apt-get update

• Clone the 'bee_track' repo:

git clone https://github.com/SheffieldMLtracking/bee_track.git

• Clone the retrodetect repo:

git clone https://github.com/SheffieldMLtracking/btretrodetect.git

• Run aravissetup from the root directory

source bee_track/aravissetup , entering "y" when prompted

This should install any dependencies needed for aravis, clone aravis, setup, build and install it ready to be used.

• THEN install python dependencies in the venv:

pip install -r bee_track/requirementsJLW250130.txt NB: this file is temporary while I test for bugs. Soon I will merge it withe "requirements.txt."

This must be done after running aravissetup because some of the python modules rely on some of the installs when installing aravis.

Before we can run btretrodetect, we need to give it an offset file so it knows how to translate between monochrom and colour camera coordinates. We will work out the true offset later, but for now we need to give it a dummy file so it doesn't crash on boot: -Create the necessary directory:

mkdir /home/pi/.btretrodetect

• Create a blank offset file:

sudo nano /home/pi/.btretrodetect/offset.json

• Now paste the following text and save the file:

{"all": [0, 0]}

• Set the device ID file: make a new file called 'device_id.txt' in the ~/bee_track/webinterface folder:

sudo nano ~/bee_track/webinterface/device_id.txt

In the new file write a single number identifying the box, on a single line line (e.g. 1015).

Running Beetrack on Boot automatically

To run the project on a headless pi with no interaction you must complete the following steps. Bear in mind it takes a good 3-5 minutes to get up and running every boot.

It looks as though the latest version of the Raspberry Pi OS no longer uses etc/rc.local and etc/network/interfaces, so we will reinstate them ourselves. (In the long term, we should look into a better way to do this.)

• Create rc.local file:

sudo nano /etc/rc.local

• Paste the following text into file:

#!/bin/sh
# add your commands
# call your scripts here
su - pi -c /home/pi/bee_track/startup &
/home/pi/bee_track/startupssh
# last line must be exit 0
exit 0

Note that this both starts beetrack on startup, and starts autossh to establish a connection to the AWS server (see section below). If you do not need the autossh connection, delete the line /home/pi/bee_track/startupssh

• Set permissions:

sudo chmod -v +x /etc/rc.local

• Enable rc.local:

sudo systemctl enable rc-local.service

• Download etc/network/interfaces:

sudo apt install ifupdown net-tools

• Now edit file by:

sudo nano /etc/network/interfaces

• ... add the following, then save and exit:

auto lo
iface lo inet loopback

auto eth0
iface eth0 inet static
address 169.254.160.220

• When you reboot (sudo reboot), the project should run without any interaction. You can reboot now, or continue with the following sections before rebooting.

Setting up autossh

We use autossh to establish a connection to our AWS server. This is used to allow us to access the boxes remotely and for the automatic data backup pipeline. We also use ssh connections to remotely use the web interface to control boxes.

If you need to access or backup boxes remotely, follow these steps:

• Install autossh:

sudo apt install autossh

• If you did not already do so in the previous step, edit rc.local to add the following line before exit 0:

/home/pi/bee_track/startupssh

Edit startupssh by

sudo nano /home/pi/bee_track/startupssh

• Edit the file to read:

sleep 30
autossh -f -R 5003:localhost:22 [email protected] -i/home/pi/.ssh/KEYFILE.pem -N -o StrictHostKeyChecking=no
autossh -f -R 8003:localhost:8000 [email protected] -i/home/pi/.ssh/KEYFILE.pem -N -o StrictHostKeyChecking=no
autossh -f -R 6003:localhost:5000 [email protected] -i/home/pi/.ssh/KEYFILE.pem -N -o StrictHostKeyChecking=no

Replace KEYFILE.pem with the name of the keyfile we use to ssh to the AWS server.

Replace 5003, 8003 and 6003 with port numbers that are unique to your box, so that we can establish connections to multiple boxes.

I have been using the following pattern: each box is identified by a four digit number, e.g. 1035. For port 22, add 4000, to get e.g. 5035. For port 8000, add 7000, to get e.g. 8035. For port 5000, add 5000, to get e.g. 6035

• Set permissions:

sudo chmod 755 /home/pi/bee_track/startupssh

• Ensure you are in the home directory (cd). Try to navigate to '.ssh' directory by

cd .ssh. If directory does not exist, create it by mkdir .ssh, then navigate to it by cd .ssh.

• Get key file for AWS server. I'm not going to publicly post instructions for that, so contact Mike or Joe if you need access.

• Set permissions:

chmod 400 KEYFILE.pem

Adding further wifi networks

We typically set up our Pis to automatically join a wifi hotspot to give us local network access, by specifying this when flishing with Raspberry Pi imager. You may also want the boxes to automatically join other networks if they are available. For example, our Ohio boxes join the University network.

• The simplest way to do this is via the network manager, which you call by:

sudo nmtui

Follow the GUI menu system to create a new wifi connection and specify the network name and password.

If you want to make edits

You will probably want to reinstall with git an editable version:

• First do "beetrack":

(bee-venv)$ pip uninstall bee_track

(bee-venv)$ cd ~/bee_track

(bee-venv)$ pip install -e .

• Now do the same for "btretrodetect":

(bee-venv)$ pip uninstall btretrodetect

(bee-venv)$ cd ~/btretrodetect

(bee-venv)$ pip install -e .

Temporarily, we need to make a edit to beetrack to allow it to import btretrodetect. In the future we will update bee_track so this is not necessary, but I don't want to do this while Mike is still working on btretrodetect

Edit "core.py":

sudo nano ~/bee_track/bee_track/core.py

Find the line that reads: import retrodetect as rd Replace with:

import btretrodetect as rd

Now reboot the pi: Bee_track should start automatically

sudo reboot

Running Beetrack from command line

• Make sure you are in virtual environment where all the installs have occured

source bee-venv/bin/activate

• Go to bee_track directory:

cd bee_track

• Start ethernet connection to camereas:

sudo ifconfig eth0 up 169.254.160.220

• Check cameras are working:

arv-camera-test-0.10

• Start bee_track:

./startupfast

• If something goes wrong use:

killall python3

• Connect to Pi with:

http://raspberrypi.local:8000/

Manual Setup

Python Setup

In root directory of the pi (or create project dir with all files in) git clone https://github.com/lionfish0/bee_track.git

Setup virtual environment python -m venv bee-venv

Activate - (if using SSH terminal to change python env) Make sure all commands are run in the virtual environment from now on source bee-venv/bin/activate

pip install -e bee_track/. sudo apt install libatlas3-base

Aravis setup

git clone https://github.com/AravisProject/aravis.git

sudo apt install cmake

and

sudo apt install libgtk-3-dev libnotify-dev libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev gstreamer1.0-plugins-bad

and

sudo apt-get install gnome-common intltool valac libglib2.0-dev gobject-introspection libgirepository1.0-dev libgtk-3-dev libgnome-desktop-3-dev libcanberra-dev libgdata-dev libdbus-glib-1-dev libgstreamer1.0-dev libupower-glib-dev libxml2-dev

meson setup aravis/

cd aravis

meson build

cd build

meson configure -Dviewer=enabled -Dintrospection=enabled -Dgst-plugin=enabled

sudo ninja install might have to pip install ninja

Beetrack setup

Important installs:

pip install PyGObject rpimotorlib rpi-lgpio psutil spidev requests

pip install scipy
git clone https://github.com/lionfish0/retrodetect.git
cd retrodetect
pip install -e .
git clone https://github.com/lionfish0/QueueBuffer.git
cd QueueBuffer
pip install -e .
pip install libsvm
pip install -U flask-cors
pip install mem_top
pip install flask_compress

Architecture

This is a simplified overview of the different parts of this system and how they communicate.

flowchart LR
user(("Web\nbrowser"))
backend[API\nFlask app\nbee_track/core.py]
frontend[Javascript GUI\nhttp.server\nwebinterface/]
user <--"HTTP 80" --> frontend
frontend <--"HTTP 5000"--> backend

Loading

API architecture

The Flask application has four components which run in separate threads:

• Cameras
• Triggers
• Rotation
• Tracking

Each thread has a worker process with a configuration message queue.

Usage

TODO

Development

Testing

TODO

Virtual environment

These are instructions for running the API app on a virtual Raspberry Pi machine using dockerpi.

1. Install Docker
2. Download the latest stable Raspberry Pi OS Lite image
3. Decompress the image unxz *.xz
4. Run a virtual machine (p3 means Rasp. Pi version 3, which is under experimental support)

image_path="./2023-12-11-raspios-bookworm-arm64-lite.img"
docker run -it -v $image_path:/sdcard/filesystem.img lukechilds/dockerpi:vm p3

TODO this doesn't work

Old instructions for setting up bee-track to run on boot. Saved for now but currently superceded by section above.

• Edit rc.local by

sudo nano /etc/rc.local

• Add the following line:

su - pi -c /home/pi/bee_track/startup &

Before exit 0

• Add the following to /etc/network/interfaces

auto lo
iface lo inet loopback

auto eth0
iface eth0 inet static
address 169.254.160.220

• We can create and enable rc.local ourselves, but I haven't yet figured out how to re-enable etc/network/interfaces. Instead, I am using rc.local to establish the ethernet connection for the cameras on startup:*

• Create rc.local file:

sudo nano /etc/rc.local

• Write in the following:

#!/bin/sh
# add your commands
# call your scripts here
ifconfig eth0 up 169.254.160.220
su - pi -c /home/pi/bee_track/startup &
/home/pi/bee_track/startupssh
# last line must be exit 0
exit 0

Note that this does three things: it replaces etc/network/interfaces to configure the ethernet connection; it starts beetrack on startup; and it starts autossh to establish a connection to the AWS server (see section below). If you do not need the autossh connection, delete the line /home/pi/bee_track/startupssh

• Set permissions:

sudo chmod -v +x /etc/rc.local

• Enable rc.local:

sudo systemctl enable rc-local.service

• When you reboot (sudo reboot), the project should run without any interaction. You can reboot now, or continue with the following sections before rebooting.