CSI Camera V2 +as of Sat 23 Mar 08:47:12 UTC 2024.
+Example for older operating systems (those with command raspistill):
csi-legacy.dist as .env if you want to use Raspberry Pi camera.env replace token-change-me with the value of the token
+ you copied.env replace fingerprint-change-me with some random value,
+ which is alphanumeric and has at least 16 chars (and max of 40 chars),
+ for example set it to fingerprint-myprinter-camera-1.envNext, test config.
+Some older Rpi 3 with older Debian with basic cam:
+PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=token-change-me
+PRUSA_CONNECT_CAMERA_FINGERPRINT=trash-cam-night-video-wide-1
+CAMERA_DEVICE=/dev/video0
+CAMERA_COMMAND=raspistill
+CAMERA_COMMAND_EXTRA_PARAMS="--nopreview --mode 640:480 -o"
+
CSI Camera V2 +as of Sat 23 Mar 08:47:12 UTC 2024.
+Example for newer operating systems (commands libcamera or rpicam-still):
csi.dist as .env if you want to use Raspberry Pi camera.env replace token-change-me with the value of the token
+ you copied.env replace fingerprint-change-me with some random value,
+ which is alphanumeric and has at least 16 chars (and max of 40 chars),
+ for example set it to fingerprint-myprinter-camera-1.envNext, test config.
+My Rpi Zero W with Raspberry Pi Camera v2 with +maximum resolution available:
+ +PRINTER_ADDRESS=192.168.1.25
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=c10eb887-f107-41a4-900e-2c38ea12a11c
+CAMERA_DEVICE=/dev/video0
+CAMERA_COMMAND=rpicam-still
+CAMERA_COMMAND_EXTRA_PARAMS="--immediate --nopreview --mode 2592:1944:12:P --lores-width 0 --lores-height 0 --thumb none -o"
+
With esphome camera with snapshot we can use the ultimate power of curl
+command to fetch the image from the camera.
Configure esphome device:
+esp32_camera and esp32_camera_web_server with
+ snapshot modules:esp32_camera:
+... (skipped due to the fact there are different modules)
+
+esp32_camera_web_server:
+ - port: 8081
+ mode: snapshot
+Flash the device and wait until it boots and is available.
+esphome-snapshot.dist as .env.env replace token-change-me with the value
+ of the token you copied.env replace fingerprint-change-me with some
+ random value, which is alphanumeric and has at least 16 chars (and max of 40 chars),
+ for example set it to fingerprint-myprinter3-camera-3.env replace your esphome device address and port
+ in CAMERA_COMMAND_EXTRA_PARAMS.envNext, test config.
+I have esp32-wrover-dev board with camera + esphome + web ui for camera exposing
+snapshot frame on port 8081.
We can use curl to fetch it.
+PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=06f47777-f179-4025-bd80-9e4cb8db2aed
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=curl
+CAMERA_COMMAND_EXTRA_PARAMS=http://esp32-wrover-0461c8.local:8081/ -o
+
With esphome camera stream we can use the ffmpeg to fetch the image from the
+camera stream. It requires a bit more computing power from esp device and the
+host that runs the image processing.
Notice that this is not recommended way due to the amount of consumed resources.
+Configure esphome device:
+esp32_camera and esp32_camera_web_server with
+ stream modules:esp32_camera:
+... (skipped due to the fact there are different modules)
+
+esp32_camera_web_server:
+ - port: 8080
+ mode: stream
+Flash the device and wait until it boots and is available.
+esphome-stream.dist as .env.env replace token-change-me with the value
+ of the token you copied.env replace fingerprint-change-me with some
+ random value, which is alphanumeric and has at least 16 chars (and max of 40 chars),
+ for example set it to fingerprint-myprinter3-camera-3.env replace your esphome device address and port
+ in CAMERA_COMMAND_EXTRA_PARAMS-update 1 may not be needed in certain ffmpeg versions.envNext, test config.
+The same ESP device with stream, notice different port (8080).
PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=token-change-me
+PRUSA_CONNECT_CAMERA_FINGERPRINT=f68336b-8dab-42cd-8729-6abd8855ff63
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=ffmpeg
+CAMERA_COMMAND_EXTRA_PARAMS="-y -i 'http://esp32-wrover-0461c8.local:8080/' -vframes 1 -q:v 1 -f image2 -update 1 "
+This processing requires ffmpeg package.
+Most standalone webcams are actually mjpg cams, they send infinite motion jpeg stream +over specific URL.
+The best option to check what is the URL is in the camera manual, or if you +open web UI of the camera and see the stream image then right click on the image +and select Inspect to see the URL for the image - copy that URL.
+You should be able to test the stream locally with ffplay command.
For example, if your camera is reachable over address 192.168.0.20 and port 8000
+under endpoint /ipcam/mjpeg.cgi then below command should show the stream:
There may be some user and password in the URL.
+If that works, then configuration should be pretty straightforward:
+ffmpeg-mjpg-stream.dist as .env.env replace token-change-me with the value
+ of the token you copied.env replace fingerprint-change-me
+ with some random value, which is alphanumeric and has at least 16 chars
+ (and max of 40 chars), for example set it to fingerprint-myprinter4-camera-4.env replace your RTSP device address raspberry-pi,
+ port and stream id in CAMERA_COMMAND_EXTRA_PARAMS if needed.envNext, test config.
+Beagle Camera stream - if I remember correctly, then camera url to the stream
+is something like http://192.168.2.92/ipcam/mjpeg.cgi
Replace 192.168.2.92 with your address in the example below.
PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=token-change-me
+PRUSA_CONNECT_CAMERA_FINGERPRINT=fingerprint-change-me
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=ffmpeg
+CAMERA_COMMAND_EXTRA_PARAMS="-y -i 'http://192.168.2.92/ipcam/mjpeg.cgi' -vframes 1 -q:v 1 -f image2 -update 1 "
+But it is better to use a snapshot instead of stream if available, +see here.
+ +DO NOT use VLC to test streams, there are unfortunately problems with it.
+Please use ffplay from ffmpeg package.
You have some options such as TCP or UDP stream (whatever..). +This should work with any other camera (usually there is a different port per stream)
+You should be able to test the stream locally with ffplay command.
For example, if your camera is reachable over address 192.168.0.20 and port 8000
+under endpoint /stream then below command should show the stream:
If that works, then configuration should be pretty straightforward:
+ffmpeg-mediamtx-rtsp-tcp.dist as .env.env replace token-change-me with the value
+ of the token you copied.env replace fingerprint-change-me
+ with some random value, which is alphanumeric and has at least 16 chars
+ (and max of 40 chars), for example set it to fingerprint-myprinter4-camera-4.env replace your RTSP device address raspberry-pi,
+ port and stream id in CAMERA_COMMAND_EXTRA_PARAMS if needed.envYou can try with UDP, but you may not get it ;-)
Next, test config.
+My another Rpi Zero W named hormex has two cameras:
and I'm running mediamtx server to conver those to RTSP streams.
+More about mediamtx is here.
So I can have two configs:
+.stream-csi over UDP:
PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=62e8ab72-9766-4ad5-b8b1-174d389fc0d3
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=ffmpeg
+CAMERA_COMMAND_EXTRA_PARAMS="-loglevel error -y -rtsp_transport udp -i "rtsp://hormex:8554/cam" -f image2 -vframes 1 -pix_fmt yuvj420p "
+.stream-endo over TCP:
PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=01a67af8-86a3-45c7-b6e2-39e9d086b367
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=ffmpeg
+CAMERA_COMMAND_EXTRA_PARAMS="-loglevel error -y -rtsp_transport tcp -i "rtsp://hormex:8554/endoscope" -f image2 -vframes 1 -pix_fmt yuvj420p "
+Some cameras expose single image snapshot under specific URL.
+we can use the ultimate power of curl command to fetch the image from the camera.
This is the preferred way to use web cams because right now Prusa Connect do not +support streams, and thus there is no point in wasting CPU on that.
+The best option to check what is the URL is in the camera manual, or if you +open web UI of the camera and see the still image then right click on the image +and select Inspect to see the URL for the image - copy that URL.
+You should be able to test the stream locally with ffplay command.
For example, if your camera is reachable over address 192.168.0.20 and port 8001
+under endpoint /snap.jpg then below command should show the image:
then you should see in the output something like Content-Type: image/jpeg,
+then you are good - see snap.jpg in the folder you executed the command.
snapshot.dist as .env.env replace token-change-me with the value
+ of the token you copied.env replace fingerprint-change-me with some
+ random value, which is alphanumeric and has at least 16 chars (and max of 40 chars),
+ for example set it to fingerprint-myprinter3-camera-3.env replace your esphome device address and port
+ in CAMERA_COMMAND_EXTRA_PARAMS.envNext, test config.
+For more in-depth details see esphome snapshot.
+I have esp32-wrover-dev board with camera + esphome + web ui for camera exposing
+snapshot frame on port 8081.
We can use curl to fetch it.
+PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=06f47777-f179-4025-bd80-9e4cb8db2aed
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=curl
+CAMERA_COMMAND_EXTRA_PARAMS=http://esp32-wrover-0461c8.local:8081/ -o
+This is not tested, I do not own such camera so hard to tell if this is right.
+Camera URL for snapshot http://192.168.2.92/images/snapshot0.jpg so the config
+should be like below:
PRINTER_ADDRESS=127.0.0.1
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=06f47777-f179-4025-bd80-9e4cb8db2aed
+CAMERA_DEVICE=/dev/null
+CAMERA_COMMAND=curl
+CAMERA_COMMAND_EXTRA_PARAMS=http://192.168.2.92/images/snapshot0.jpg -o
+
This should work on any linux distro with any sane camera that you have.
+Run v4l2-ctl --list-devices.
This should show list of devices to use, where /dev/video0 is a device
+name.
Notice that not every device is an actual camera.
+The quick all-in one output for camera /dev/video0 is
For more details about formats it is better to use
+v4l2-ctl --list-formats-ext -d /dev/video0
usb.dist as .env.env replace token-change-me with the value of the token
+ you copied.env replace fingerprint-change-me with some random value,
+ which is alphanumeric and has at least 16 chars (and max of 40 chars),
+ for example set it to fingerprint-myprinter2-camera-2.env replace /dev/video0 with desired device in CAMERA_DEVICE.envNext, test config.
+Raspberry Pi Zero W with endoscope camera over USB, registered as /dev/video1:
PRINTER_ADDRESS=192.168.1.25
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=7054ba85-bc19-4eb9-badc-6129575d9651
+CAMERA_DEVICE=/dev/video1
+CAMERA_COMMAND=fswebcam
+CAMERA_COMMAND_EXTRA_PARAMS="--resolution 1280x960 --no-banner"
+Microsoft LifeCam HD-3000, USB, to avoid image capture tearing we skip first 10
+frames -S 10 which also helps to set up proper auto exposure and white balance.
PRINTER_ADDRESS=192.168.1.25
+PRUSA_CONNECT_CAMERA_TOKEN=redacted
+PRUSA_CONNECT_CAMERA_FINGERPRINT=7054ba85-bc19-4eb9-badc-6129575d9651
+CAMERA_DEVICE=/dev/video4
+CAMERA_COMMAND=fswebcam
+CAMERA_COMMAND_EXTRA_PARAMS="-S 10 --resolution 1280x720 --no-banner -s auto_exposure=1,brightness=128,contrast=5"
+PRUSA_CONNECT_CAMERA_TOKEN should be taken from earlier step.
PRUSA_CONNECT_CAMERA_FINGERPRINT should be uniqe and set only once for each camera.
Fingerprint can be easily generated using command:
+ +or via online website, +just copy/paste the output as fingerprint value into the config.
+Do not change fingerprint after launching the script - thus camera is registered +and you may need to revert the change or delete and readd camera again and start +from scratch.
+Other env vars are set depending on the camera device we want to use.
+Next, test config.
+ +Config for camera is to the script as environment variables (env vars).
+SLEEP - sleep time in seconds between image captures,
+ notice that PrusaConnect accepts images at most every 10s or slower.
+ Default value 10.
PRINTER_ADDRESS - Printer address to ping, if address is unreachable there
+ is no point in sending an image. Set to 127.0.0.1 to always send images.
+ Set to empty value to disable ping check and always send images.
+ Default value 127.0.0.1
PRUSA_CONNECT_CAMERA_TOKEN - required, PrusaConnect API key
PRUSA_CONNECT_CAMERA_FINGERPRINT - required, PrusaConnect camera fingerprint,
+ use for example cli uuidgen or web
+ to generate it, it must be at least 16 alphanumeric chars, 40 max.
+ Remember not to change this if it was already set, otherwise you need to
+ remove and add the camera again.
CAMERA_DEVICE - camera device to use, if you use Raspberry Pi camera
+ attached to the CSI via camera ribbon then leave as is
+ Default /dev/video0 which points to first detected camera.
CAMERA_SETUP_COMMAND - camera setup command and params executed before
+ taking image, default value is empty, because some cameras do not support it,
+ in general you want to use something like v4l2-ctl parameters, so
+ so for example
+ setup_command=v4l2-ctl --set-ctrl brightness=10,gamma=120 -d $CAMERA_DEVICE
+ will translate to:
+ v4l2-ctl --set-ctrl brightness=10,gamma=120 -d /dev/video0
CAMERA_COMMAND - command used to invoke image capture,
+ default is rpicam-still
+ available options:
anything else will be processed directly, so for example you could use + 'ffmpeg' in here
+CAMERA_COMMAND_EXTRA_PARAMS -extra params passed to the camera program,
+ passed directly as <command> <extra-params> <output_file>
+ example values per specific camera:
+
+
libcamera (rpicam-still)
+ --immediate --nopreview --mode 2592:1944:12:P --lores-width 0 --lores-height 0 --thumb none -o
--nopreview --mode 2592:1944:12:P -o--resolution 1280x960 --no-banner-f v4l2 -y -i /dev/video0 -f image2 -vframes 1 -pix_fmt yuvj420pTARGET_DIR - directory where to save camera images, image per camera will
+ be overwritten per image capture,
+ default value /dev/shm so that we do not write to microSD cards or read only
+ filesystems/containers. /dev/shm is a shared memory space. if you have more
+ printers you may need to increase this value on system level.
CURL_EXTRA_PARAMS - extra params to curl when pushing an image,
+ default empty value, but you could for example add additional params if needed
+ such as -k if using tls proxy with self-signed certificate
PRUSA_CONNECT_URL - Prusa Connect endpoint where to post images,
+ default value https://webcam.connect.prusa3d.com/c/snapshot
+ You could put here Prusa Connect Proxy if you use one.
For more in-depth details (no need to repeat them here) please see the top of +the prusa-connect-camera.sh.
+ +Config for camera is to the script as environment variables (env vars).
+The most important env vars are:
+PRUSA_CONNECT_CAMERA_TOKENPRUSA_CONNECT_CAMERA_FINGERPRINTCAMERA_COMMANDCAMERA_COMMAND_EXTRA_PARAMSThose env vars will be filled in in the next steps.
+Full list of env vars can be seen here
+ +Short overview of actions:
+Assuming you already have a working camera with basic setup, we can tune it further.
+Below steps depend on the camera capabilities, thus your mileage may vary.
+Notice that Prusa Connect has file size limit something about 8MB of the image uploaded, +so there may be no point in getting images with super high resolutions.
+Use v4l2-ctl to get the list of available resolutions that camera provides
+and then update it in the env var configs.
Run v4l2-ctl --list-formats-ext -d /dev/video0 where /dev/video0 is a device
+listed from command above.
Example output:
+v4l2-ctl --list-formats-ext -d /dev/video1
+ioctl: VIDIOC_ENUM_FMT
+ Type: Video Capture
+
+ [0]: 'MJPG' (Motion-JPEG, compressed)
+ Size: Discrete 640x480
+ Interval: Discrete 0.033s (30.000 fps)
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 640x360
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 352x288
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 320x240
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 176x144
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 160x120
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 800x600
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 1280x720
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 1280x960
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 640x480
+ Interval: Discrete 0.033s (30.000 fps)
+ Interval: Discrete 0.033s (30.000 fps)
+ [1]: 'YUYV' (YUYV 4:2:2)
+ Size: Discrete 640x480
+ Interval: Discrete 0.033s (30.000 fps)
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 640x360
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 352x288
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 320x240
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 176x144
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 160x120
+ Interval: Discrete 0.033s (30.000 fps)
+ Size: Discrete 800x600
+ Interval: Discrete 0.200s (5.000 fps)
+ Size: Discrete 1280x720
+ Interval: Discrete 0.200s (5.000 fps)
+ Size: Discrete 1280x960
+ Interval: Discrete 0.200s (5.000 fps)
+ Size: Discrete 640x480
+ Interval: Discrete 0.033s (30.000 fps)
+ Interval: Discrete 0.033s (30.000 fps)
+As you can see if I set video to YUYV and with resolution higher than 800x600 +I would get only 5 frames per second. +For still images this is not a problem, but for video streaming that could be +too low and I would have to switch to MJPG (or actually mjpeg in ffmpeg)
+For Raspberry Cam v2 you could use csi.dist as source and add
+--mode 2592:1944:12:P to the CAMERA_COMMAND_EXTRA_PARAMS.
For certain USB cameras (such as Tracer Endoscope) you should use usb.dist and
+you should be able to add --resolution 1280x960 to the CAMERA_COMMAND_EXTRA_PARAMS.
Video controls are things like brightness, auto white balance (awb), +exposure and so on.
+Get device capabilities, especially User controls:
Those params can be passed in various ways, depending on the requirement:
+CAMERA_COMMAND_EXTRA_PARAMS emv var, when using given tool directlyCAMERA_SETUP_COMMAND env var for some complex use casesand set accordingly parameters you want , you just need to pass them fswebcam, +for example:
+CAMERA_COMMAND_EXTRA_PARAMS=--resolution 1280x960 --no-banner -s auto_exposure=1,brightness=128,contrast=5
For more advanced options see CAMERA_SETUP_COMMAND env var, for example:
remember to restart given camera service.
+You can try to use guvcview desktop application to check prams in realtime.
If your captured image has below issues:
+the whole image is too dark or too bright and it changes with every capture
+ so it get too dark or too bright in a matter of minutes:
+ 
+ 
+ 
has some horizontal/vertical super bright/dark areas:
+ 
+ 
some visible artifacts such as colored blocks or missing image fragments:
+ 
+ 
+ 
then you may need to initialize camera and capture it with a delay +or drop initial number of frames.
+Notice that sometimes you cannot do much about it (remote cams) because some +camera images will be broken anyway, then I suggest changing camera.
+Usually this is the fastest and with fswebcam it can be achieved by passing -S
+param, for example
-S 10 will skip 10 first frames.
fswebcam param -d 2 will delay capture for 2 seconds, for some cameras it may
+help.
You can pass on params to rpicam-still or fswebcam as you want.
+See rpicam-still --help
--hflip Read out with horizontal mirror
+ --vflip Read out with vertical flip
+ --rotation Use hflip and vflip to create the given rotation <angle>
+so for example:
+CAMERA_COMMAND=rpicam-still
+CAMERA_COMMAND_EXTRA_PARAMS="--rotation 90 --immediate --nopreview --thumb none -o"
+See fswebcam --help
--flip <direction> Flips the image. (h, v)
+ --crop <size>[,<offset>] Crop a part of the image.
+ --scale <size> Scales the image.
+ --rotate <angle> Rotates the image in right angles.
+so for example:
+ +or to skip first 10 frames (-S 10, helps to get proper auto white balance and
+image exposure) and pass on camera controls:
CAMERA_COMMAND=fswebcam
+CAMERA_COMMAND_EXTRA_PARAMS="-S 10 --resolution 1280x720 --no-banner -s auto_exposure=1,brightness=128,contrast=5"
+When curl is not enough and you don't really want to physically rotate your camera, +then use ffmpeg for post processing. +You can process static images with it, load v4l2 devices... whatever.
+With ffmpeg you can do interesting things with filters, it will just require +more computing power.
+v4l2 can be used as alias for video4linux2.
You can pass video4linux options to ffmpeg on device initialization, for example:
+ +ffmpeg -f v4l2 -pix_fmt mjpeg -video_size 1280x960 -framerate 30 -i /dev/video1 \
+ -c:v libx264 -preset ultrafast -b:v 6000k -f rtsp rtsp://localhost:$RTSP_PORT/$MTX_PATH
+would instruct ffmpeg to use video4linux and force it to talk to the camera under +/dev/video1 and forcing mjpeg encoder, resolution and framerate.
+This command above is directly taken from mediamtx.
+For more params, see official ffmpeg docs.
+Just remember to pass them before defining input (-i /dev/video1).
See here +for basic ones.
+You probably want to use -vf "transpose=1" to rotate image 90 degrees clockwise:
CAMERA_COMMAND=ffmpeg
+CAMERA_COMMAND_EXTRA_PARAMS="-y -i 'http://esp32-wrover-0461c8.local:8080/' -vf 'transpose=1' -vframes 1 -q:v 1 -f image2 -update 1 "
+Frankly speaking you can do anything you want with ffmpeg, for example
+-vf transpose=1,shufflepixels=m=block:height=16:width=16
Why? why not :D
+ +
This project aims to make it easier to use any camera to be used as +Prusa Connect camera.
+Rpi Zero W or older devices may have CPU limitations to process remote streams + or multiple cameras at once
+I was not able to test EVERY setting so this may still have some bugs
+Install system packages - assuming Debian based distros on Raspberry Pi OS, which +also come in with some pre-installed packages.
+Below commands should be executed in shell/terminal (on the Raspberry Pi).
+For most Raspberry Pi Cameras (CSI/USB):
+sudo apt-get update
+sudo apt-get install -y curl libcamera0 fswebcam git iputils-ping v4l-utils uuid-runtime
+Additional packages for remote cameras - especially the one that are used for streaming:
+ +Download this script:
+mkdir -p /home/pi/src
+cd /home/pi/src
+git clone https://github.com/nvtkaszpir/prusa-connect-camera-script.git
+cd prusa-connect-camera-script
+Raspberry Pi Zero W is able to process CSI camera + (Rpi Cam v2) and USB 2k camera + but it has load average about 1.4, and CPU is quite well utilized, so you may + need to decrease resolution per camera to see how + it goes.
+for webcams it is always better to choose snapshot + because it requires less computing both on camera and on the host, + otherwise we need to use ffmpeg
+ffmpeg is usually noticeably slow and cpu intensive, especially if you do more + complex operations
+PrinterCameraAdd new other cameraToken, this is needed later as
+ PRUSA_CONNECT_CAMERA_TOKEN env varPhysical host or virtual machine or container:
+Camera such as:
+ +esp32_camera_web_server with snapshot moduleesp32_camera_web_server with stream module using ffmpegffmpegLinux operating system. +Debian based preferred, for example Raspberry Pi OS Lite if you run Raspberry Pi. +I use also laptop with Ubuntu 22.04, but I believe with minor tweaks it should +work on most distributions (mainly package names are different).
+Below list uses Debian package names.
+bash 5.x (what year is it?)git (just to install scripts from this repo)curliputils-pinguuid-runtime to make generation of camera fingerprint easierv4l-utils - to detect camera capabilitieslibcamera0 - for Rpi CSI cameraslibraspberrypi-bin or rpicam-apps-lite for Rpi CSI cameras
+ (should be already installed on Rpi OS)fswebcam - for generic USB camerasffmpeg - for custom commands for capturing remote streamsThis project aims to make it easier to use any camera to be used as Prusa Connect camera.
"},{"location":"#features","title":"Features","text":"Rpi Zero W or older devices may have CPU limitations to process remote streams or multiple cameras at once
I was not able to test EVERY setting so this may still have some bugs
CSI Camera V2 as of Sat 23 Mar 08:47:12 UTC 2024.
Example for older operating systems (those with command raspistill):
csi-legacy.dist as .env if you want to use Raspberry Pi camera.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter-camera-1.envNext, test config.
"},{"location":"config.for.camera.csi.legacy/#real-world-scenario","title":"Real world scenario","text":"Some older Rpi 3 with older Debian with basic cam:
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=token-change-me\nPRUSA_CONNECT_CAMERA_FINGERPRINT=trash-cam-night-video-wide-1\nCAMERA_DEVICE=/dev/video0\nCAMERA_COMMAND=raspistill\nCAMERA_COMMAND_EXTRA_PARAMS=\"--nopreview --mode 640:480 -o\"\nCSI Camera V2 as of Sat 23 Mar 08:47:12 UTC 2024.
Example for newer operating systems (commands libcamera or rpicam-still):
csi.dist as .env if you want to use Raspberry Pi camera.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter-camera-1.envNext, test config.
"},{"location":"config.for.camera.csi.libcamera/#real-example","title":"Real example","text":"My Rpi Zero W with Raspberry Pi Camera v2 with maximum resolution available:
PRINTER_ADDRESS=192.168.1.25\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=c10eb887-f107-41a4-900e-2c38ea12a11c\nCAMERA_DEVICE=/dev/video0\nCAMERA_COMMAND=rpicam-still\nCAMERA_COMMAND_EXTRA_PARAMS=\"--immediate --nopreview --mode 2592:1944:12:P --lores-width 0 --lores-height 0 --thumb none -o\"\nWith esphome camera with snapshot we can use the ultimate power of curl command to fetch the image from the camera.
Configure esphome device:
esp32_camera and esp32_camera_web_server with snapshot modules:esp32_camera:\n... (skipped due to the fact there are different modules)\n\nesp32_camera_web_server:\n - port: 8081\n mode: snapshot\nFlash the device and wait until it boots and is available.
"},{"location":"config.for.camera.esphome.snapshot/#create-config-for-script","title":"Create config for script","text":"esphome-snapshot.dist as .env.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter3-camera-3.env replace your esphome device address and port in CAMERA_COMMAND_EXTRA_PARAMS.envNext, test config.
"},{"location":"config.for.camera.esphome.snapshot/#real-world-example","title":"Real world example","text":"I have esp32-wrover-dev board with camera + esphome + web ui for camera exposing snapshot frame on port 8081.
We can use curl to fetch it.
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=06f47777-f179-4025-bd80-9e4cb8db2aed\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=curl\nCAMERA_COMMAND_EXTRA_PARAMS=http://esp32-wrover-0461c8.local:8081/ -o\nWith esphome camera stream we can use the ffmpeg to fetch the image from the camera stream. It requires a bit more computing power from esp device and the host that runs the image processing.
Notice that this is not recommended way due to the amount of consumed resources.
"},{"location":"config.for.camera.esphome.stream/#prepare-esphome-device","title":"Prepare esphome device","text":"Configure esphome device:
esp32_camera and esp32_camera_web_server with stream modules:esp32_camera:\n... (skipped due to the fact there are different modules)\n\nesp32_camera_web_server:\n - port: 8080\n mode: stream\nFlash the device and wait until it boots and is available.
"},{"location":"config.for.camera.esphome.stream/#create-config-for-script","title":"Create config for script","text":"esphome-stream.dist as .env.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter3-camera-3.env replace your esphome device address and port in CAMERA_COMMAND_EXTRA_PARAMS-update 1 may not be needed in certain ffmpeg versions.envNext, test config.
"},{"location":"config.for.camera.esphome.stream/#real-world-example","title":"Real world example","text":"The same ESP device with stream, notice different port (8080).
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=token-change-me\nPRUSA_CONNECT_CAMERA_FINGERPRINT=f68336b-8dab-42cd-8729-6abd8855ff63\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=ffmpeg\nCAMERA_COMMAND_EXTRA_PARAMS=\"-y -i 'http://esp32-wrover-0461c8.local:8080/' -vframes 1 -q:v 1 -f image2 -update 1 \"\nPRUSA_CONNECT_CAMERA_TOKEN should be taken from earlier step.
PRUSA_CONNECT_CAMERA_FINGERPRINT should be uniqe and set only once for each camera.
Fingerprint can be easily generated using command:
uuidgen\nor via online website, just copy/paste the output as fingerprint value into the config.
Do not change fingerprint after launching the script - thus camera is registered and you may need to revert the change or delete and readd camera again and start from scratch.
"},{"location":"config.for.camera/#example-devices","title":"Example devices","text":"Other env vars are set depending on the camera device we want to use.
"},{"location":"config.for.camera/#locally-connected","title":"Locally connected","text":"Next, test config.
"},{"location":"config.for.camera.mjpg/","title":"Web Cam - MJPG stream","text":"This processing requires ffmpeg package.
Most standalone webcams are actually mjpg cams, they send infinite motion jpeg stream over specific URL.
The best option to check what is the URL is in the camera manual, or if you open web UI of the camera and see the stream image then right click on the image and select Inspect to see the URL for the image - copy that URL.
You should be able to test the stream locally with ffplay command.
For example, if your camera is reachable over address 192.168.0.20 and port 8000 under endpoint /ipcam/mjpeg.cgi then below command should show the stream:
ffplay http://192.168.0.20:8000/ipcam/mjpeg.cgi\nThere may be some user and password in the URL.
If that works, then configuration should be pretty straightforward:
ffmpeg-mjpg-stream.dist as .env.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter4-camera-4.env replace your RTSP device address raspberry-pi, port and stream id in CAMERA_COMMAND_EXTRA_PARAMS if needed.envNext, test config.
"},{"location":"config.for.camera.mjpg/#unverified-example","title":"Unverified example","text":"Beagle Camera stream - if I remember correctly, then camera url to the stream is something like http://192.168.2.92/ipcam/mjpeg.cgi
Replace 192.168.2.92 with your address in the example below.
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=token-change-me\nPRUSA_CONNECT_CAMERA_FINGERPRINT=fingerprint-change-me\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=ffmpeg\nCAMERA_COMMAND_EXTRA_PARAMS=\"-y -i 'http://192.168.2.92/ipcam/mjpeg.cgi' -vframes 1 -q:v 1 -f image2 -update 1 \"\nBut it is better to use a snapshot instead of stream if available, see here.
"},{"location":"config.for.camera.rtsp/","title":"Web Cam - RTSP stream","text":""},{"location":"config.for.camera.rtsp/#caution","title":"Caution","text":"DO NOT use VLC to test streams, there are unfortunately problems with it. Please use ffplay from ffmpeg package.
You have some options such as TCP or UDP stream (whatever..). This should work with any other camera (usually there is a different port per stream)
You should be able to test the stream locally with ffplay command.
For example, if your camera is reachable over address 192.168.0.20 and port 8000 under endpoint /stream then below command should show the stream:
ffplay rtsp://192.168.0.20:8000/stream\nIf that works, then configuration should be pretty straightforward:
ffmpeg-mediamtx-rtsp-tcp.dist as .env.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter4-camera-4.env replace your RTSP device address raspberry-pi, port and stream id in CAMERA_COMMAND_EXTRA_PARAMS if needed.envYou can try with UDP, but you may not get it ;-)
Next, test config.
"},{"location":"config.for.camera.rtsp/#real-world-example","title":"Real world example","text":"My another Rpi Zero W named hormex has two cameras:
and I'm running mediamtx server to conver those to RTSP streams. More about mediamtx is here.
So I can have two configs:
.stream-csi over UDP:
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=62e8ab72-9766-4ad5-b8b1-174d389fc0d3\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=ffmpeg\nCAMERA_COMMAND_EXTRA_PARAMS=\"-loglevel error -y -rtsp_transport udp -i \"rtsp://hormex:8554/cam\" -f image2 -vframes 1 -pix_fmt yuvj420p \"\n.stream-endo over TCP:
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=01a67af8-86a3-45c7-b6e2-39e9d086b367\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=ffmpeg\nCAMERA_COMMAND_EXTRA_PARAMS=\"-loglevel error -y -rtsp_transport tcp -i \"rtsp://hormex:8554/endoscope\" -f image2 -vframes 1 -pix_fmt yuvj420p \"\nSome cameras expose single image snapshot under specific URL. we can use the ultimate power of curl command to fetch the image from the camera.
This is the preferred way to use web cams because right now Prusa Connect do not support streams, and thus there is no point in wasting CPU on that.
The best option to check what is the URL is in the camera manual, or if you open web UI of the camera and see the still image then right click on the image and select Inspect to see the URL for the image - copy that URL.
You should be able to test the stream locally with ffplay command.
For example, if your camera is reachable over address 192.168.0.20 and port 8001 under endpoint /snap.jpg then below command should show the image:
curl -vvv http://another-cam.local:8081/snap.jpg -o snap.jpg\nthen you should see in the output something like Content-Type: image/jpeg, then you are good - see snap.jpg in the folder you executed the command.
snapshot.dist as .env.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter3-camera-3.env replace your esphome device address and port in CAMERA_COMMAND_EXTRA_PARAMS.envNext, test config.
"},{"location":"config.for.camera.snapshot/#real-world-example","title":"Real world example","text":""},{"location":"config.for.camera.snapshot/#esp32-with-esphome","title":"esp32 with esphome","text":"For more in-depth details see esphome snapshot.
I have esp32-wrover-dev board with camera + esphome + web ui for camera exposing snapshot frame on port 8081.
We can use curl to fetch it.
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=06f47777-f179-4025-bd80-9e4cb8db2aed\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=curl\nCAMERA_COMMAND_EXTRA_PARAMS=http://esp32-wrover-0461c8.local:8081/ -o\nThis is not tested, I do not own such camera so hard to tell if this is right.
Camera URL for snapshot http://192.168.2.92/images/snapshot0.jpg so the config should be like below:
PRINTER_ADDRESS=127.0.0.1\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=06f47777-f179-4025-bd80-9e4cb8db2aed\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=curl\nCAMERA_COMMAND_EXTRA_PARAMS=http://192.168.2.92/images/snapshot0.jpg -o\nThis should work on any linux distro with any sane camera that you have.
"},{"location":"config.for.camera.usb/#how-to-get-info-which-cameras-are-available","title":"How to get info which cameras are available?","text":"Run v4l2-ctl --list-devices.
This should show list of devices to use, where /dev/video0 is a device name.
Notice that not every device is an actual camera.
"},{"location":"config.for.camera.usb/#how-to-get-what-modes-are-available-for-the-camera","title":"How to get what modes are available for the camera?","text":"The quick all-in one output for camera /dev/video0 is
v4l2-ctl -d /dev/video0 --all\nFor more details about formats it is better to use v4l2-ctl --list-formats-ext -d /dev/video0
usb.dist as .env.env replace token-change-me with the value of the token you copied.env replace fingerprint-change-me with some random value, which is alphanumeric and has at least 16 chars (and max of 40 chars), for example set it to fingerprint-myprinter2-camera-2.env replace /dev/video0 with desired device in CAMERA_DEVICE.envNext, test config.
"},{"location":"config.for.camera.usb/#real-world-example","title":"Real world example","text":"Raspberry Pi Zero W with endoscope camera over USB, registered as /dev/video1:
PRINTER_ADDRESS=192.168.1.25\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=7054ba85-bc19-4eb9-badc-6129575d9651\nCAMERA_DEVICE=/dev/video1\nCAMERA_COMMAND=fswebcam\nCAMERA_COMMAND_EXTRA_PARAMS=\"--resolution 1280x960 --no-banner\"\nMicrosoft LifeCam HD-3000, USB, to avoid image capture tearing we skip first 10 frames -S 10 which also helps to set up proper auto exposure and white balance.
PRINTER_ADDRESS=192.168.1.25\nPRUSA_CONNECT_CAMERA_TOKEN=redacted\nPRUSA_CONNECT_CAMERA_FINGERPRINT=7054ba85-bc19-4eb9-badc-6129575d9651\nCAMERA_DEVICE=/dev/video4\nCAMERA_COMMAND=fswebcam\nCAMERA_COMMAND_EXTRA_PARAMS=\"-S 10 --resolution 1280x720 --no-banner -s auto_exposure=1,brightness=128,contrast=5\"\nConfig for camera is to the script as environment variables (env vars).
SLEEP - sleep time in seconds between image captures, notice that PrusaConnect accepts images at most every 10s or slower. Default value 10.
PRINTER_ADDRESS - Printer address to ping, if address is unreachable there is no point in sending an image. Set to 127.0.0.1 to always send images. Set to empty value to disable ping check and always send images. Default value 127.0.0.1
PRUSA_CONNECT_CAMERA_TOKEN - required, PrusaConnect API key
PRUSA_CONNECT_CAMERA_FINGERPRINT - required, PrusaConnect camera fingerprint, use for example cli uuidgen or web to generate it, it must be at least 16 alphanumeric chars, 40 max. Remember not to change this if it was already set, otherwise you need to remove and add the camera again.
CAMERA_DEVICE - camera device to use, if you use Raspberry Pi camera attached to the CSI via camera ribbon then leave as is Default /dev/video0 which points to first detected camera.
CAMERA_SETUP_COMMAND - camera setup command and params executed before taking image, default value is empty, because some cameras do not support it, in general you want to use something like v4l2-ctl parameters, so so for example setup_command=v4l2-ctl --set-ctrl brightness=10,gamma=120 -d $CAMERA_DEVICE will translate to: v4l2-ctl --set-ctrl brightness=10,gamma=120 -d /dev/video0
CAMERA_COMMAND - command used to invoke image capture, default is rpicam-still available options:
anything else will be processed directly, so for example you could use 'ffmpeg' in here
CAMERA_COMMAND_EXTRA_PARAMS -extra params passed to the camera program, passed directly as <command> <extra-params> <output_file> example values per specific camera:
libcamera (rpicam-still) --immediate --nopreview --mode 2592:1944:12:P --lores-width 0 --lores-height 0 --thumb none -o
--nopreview --mode 2592:1944:12:P -o--resolution 1280x960 --no-banner-f v4l2 -y -i /dev/video0 -f image2 -vframes 1 -pix_fmt yuvj420pTARGET_DIR - directory where to save camera images, image per camera will be overwritten per image capture, default value /dev/shm so that we do not write to microSD cards or read only filesystems/containers. /dev/shm is a shared memory space. if you have more printers you may need to increase this value on system level.
CURL_EXTRA_PARAMS - extra params to curl when pushing an image, default empty value, but you could for example add additional params if needed such as -k if using tls proxy with self-signed certificate
PRUSA_CONNECT_URL - Prusa Connect endpoint where to post images, default value https://webcam.connect.prusa3d.com/c/snapshot You could put here Prusa Connect Proxy if you use one.
For more in-depth details (no need to repeat them here) please see the top of the prusa-connect-camera.sh.
"},{"location":"configuration.env/","title":"Configuration Env Vars","text":""},{"location":"configuration.env/#minimum-required-env-vars","title":"Minimum required env vars","text":"Config for camera is to the script as environment variables (env vars).
The most important env vars are:
PRUSA_CONNECT_CAMERA_TOKENPRUSA_CONNECT_CAMERA_FINGERPRINTCAMERA_COMMANDCAMERA_COMMAND_EXTRA_PARAMSThose env vars will be filled in in the next steps.
Full list of env vars can be seen here
"},{"location":"configuration.overview/","title":"Configuration Overview","text":"Short overview of actions:
Assuming you already have a working camera with basic setup, we can tune it further.
Below steps depend on the camera capabilities, thus your mileage may vary.
Notice that Prusa Connect has file size limit something about 8MB of the image uploaded, so there may be no point in getting images with super high resolutions.
"},{"location":"configuration.tuning/#getting-higher-quality-camera-images","title":"Getting higher quality camera images","text":"Use v4l2-ctl to get the list of available resolutions that camera provides and then update it in the env var configs.
Run v4l2-ctl --list-formats-ext -d /dev/video0 where /dev/video0 is a device listed from command above.
Example output:
v4l2-ctl --list-formats-ext -d /dev/video1\nioctl: VIDIOC_ENUM_FMT\n Type: Video Capture\n\n [0]: 'MJPG' (Motion-JPEG, compressed)\n Size: Discrete 640x480\n Interval: Discrete 0.033s (30.000 fps)\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 640x360\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 352x288\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 320x240\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 176x144\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 160x120\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 800x600\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 1280x720\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 1280x960\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 640x480\n Interval: Discrete 0.033s (30.000 fps)\n Interval: Discrete 0.033s (30.000 fps)\n [1]: 'YUYV' (YUYV 4:2:2)\n Size: Discrete 640x480\n Interval: Discrete 0.033s (30.000 fps)\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 640x360\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 352x288\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 320x240\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 176x144\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 160x120\n Interval: Discrete 0.033s (30.000 fps)\n Size: Discrete 800x600\n Interval: Discrete 0.200s (5.000 fps)\n Size: Discrete 1280x720\n Interval: Discrete 0.200s (5.000 fps)\n Size: Discrete 1280x960\n Interval: Discrete 0.200s (5.000 fps)\n Size: Discrete 640x480\n Interval: Discrete 0.033s (30.000 fps)\n Interval: Discrete 0.033s (30.000 fps)\nAs you can see if I set video to YUYV and with resolution higher than 800x600 I would get only 5 frames per second. For still images this is not a problem, but for video streaming that could be too low and I would have to switch to MJPG (or actually mjpeg in ffmpeg)
For Raspberry Cam v2 you could use csi.dist as source and add --mode 2592:1944:12:P to the CAMERA_COMMAND_EXTRA_PARAMS.
For certain USB cameras (such as Tracer Endoscope) you should use usb.dist and you should be able to add --resolution 1280x960 to the CAMERA_COMMAND_EXTRA_PARAMS.
Video controls are things like brightness, auto white balance (awb), exposure and so on.
Get device capabilities, especially User controls:
v4l2-ctl -d /dev/video0 -l\nThose params can be passed in various ways, depending on the requirement:
CAMERA_COMMAND_EXTRA_PARAMS emv var, when using given tool directlyCAMERA_SETUP_COMMAND env var for some complex use casesand set accordingly parameters you want , you just need to pass them fswebcam, for example:
CAMERA_COMMAND_EXTRA_PARAMS=--resolution 1280x960 --no-banner -s auto_exposure=1,brightness=128,contrast=5
For more advanced options see CAMERA_SETUP_COMMAND env var, for example:
CAMERA_SETUP_COMMAND=\"v4l2-ctl --set-ctrl brightness=64,gamma=300 -d $CAMERA_DEVICE\"\nremember to restart given camera service.
You can try to use guvcview desktop application to check prams in realtime.
If your captured image has below issues:
the whole image is too dark or too bright and it changes with every capture so it get too dark or too bright in a matter of minutes:
has some horizontal/vertical super bright/dark areas:
some visible artifacts such as colored blocks or missing image fragments:
then you may need to initialize camera and capture it with a delay or drop initial number of frames.
Notice that sometimes you cannot do much about it (remote cams) because some camera images will be broken anyway, then I suggest changing camera.
"},{"location":"configuration.tuning/#dropping-frames","title":"Dropping frames","text":"Usually this is the fastest and with fswebcam it can be achieved by passing -S param, for example
-S 10 will skip 10 first frames.
fswebcam param -d 2 will delay capture for 2 seconds, for some cameras it may help.
You can pass on params to rpicam-still or fswebcam as you want.
"},{"location":"configuration.tuning/#rpicam-still","title":"rpicam-still","text":"See rpicam-still --help
--hflip Read out with horizontal mirror\n --vflip Read out with vertical flip\n --rotation Use hflip and vflip to create the given rotation <angle>\nso for example:
CAMERA_COMMAND=rpicam-still\nCAMERA_COMMAND_EXTRA_PARAMS=\"--rotation 90 --immediate --nopreview --thumb none -o\"\nSee fswebcam --help
--flip <direction> Flips the image. (h, v)\n --crop <size>[,<offset>] Crop a part of the image.\n --scale <size> Scales the image.\n --rotate <angle> Rotates the image in right angles.\nso for example:
CAMERA_COMMAND=fswebcam\nCAMERA_COMMAND_EXTRA_PARAMS=\"--flip v --resolution 640x480 --no-banner\"\nor to skip first 10 frames (-S 10, helps to get proper auto white balance and image exposure) and pass on camera controls:
CAMERA_COMMAND=fswebcam\nCAMERA_COMMAND_EXTRA_PARAMS=\"-S 10 --resolution 1280x720 --no-banner -s auto_exposure=1,brightness=128,contrast=5\"\nWhen curl is not enough and you don't really want to physically rotate your camera, then use ffmpeg for post processing. You can process static images with it, load v4l2 devices... whatever.
With ffmpeg you can do interesting things with filters, it will just require more computing power.
"},{"location":"configuration.tuning/#adding-v4l2-options","title":"Adding v4l2 options","text":"v4l2 can be used as alias for video4linux2.
You can pass video4linux options to ffmpeg on device initialization, for example:
ffmpeg -f v4l2 -pix_fmt mjpeg -video_size 1280x960 -framerate 30 -i /dev/video1 \\\n -c:v libx264 -preset ultrafast -b:v 6000k -f rtsp rtsp://localhost:$RTSP_PORT/$MTX_PATH\nwould instruct ffmpeg to use video4linux and force it to talk to the camera under /dev/video1 and forcing mjpeg encoder, resolution and framerate.
This command above is directly taken from mediamtx.
For more params, see official ffmpeg docs. Just remember to pass them before defining input (-i /dev/video1).
See here for basic ones.
You probably want to use -vf \"transpose=1\" to rotate image 90 degrees clockwise:
CAMERA_COMMAND=ffmpeg\nCAMERA_COMMAND_EXTRA_PARAMS=\"-y -i 'http://esp32-wrover-0461c8.local:8080/' -vf 'transpose=1' -vframes 1 -q:v 1 -f image2 -update 1 \"\nFrankly speaking you can do anything you want with ffmpeg, for example
-vf transpose=1,shufflepixels=m=block:height=16:width=16
Why? why not :D
"},{"location":"installation/","title":"Installation","text":"Install system packages - assuming Debian based distros on Raspberry Pi OS, which also come in with some pre-installed packages.
Below commands should be executed in shell/terminal (on the Raspberry Pi).
For most Raspberry Pi Cameras (CSI/USB):
sudo apt-get update\nsudo apt-get install -y curl libcamera0 fswebcam git iputils-ping v4l-utils uuid-runtime\nAdditional packages for remote cameras - especially the one that are used for streaming:
sudo apt-get install -y ffmpeg\nDownload this script:
mkdir -p /home/pi/src\ncd /home/pi/src\ngit clone https://github.com/nvtkaszpir/prusa-connect-camera-script.git\ncd prusa-connect-camera-script\nRaspberry Pi Zero W is able to process CSI camera (Rpi Cam v2) and USB 2k camera but it has load average about 1.4, and CPU is quite well utilized, so you may need to decrease resolution per camera to see how it goes.
for webcams it is always better to choose snapshot because it requires less computing both on camera and on the host, otherwise we need to use ffmpeg
ffmpeg is usually noticeably slow and cpu intensive, especially if you do more complex operations
PrinterCameraAdd new other cameraToken, this is needed later as PRUSA_CONNECT_CAMERA_TOKEN env varPhysical host or virtual machine or container:
Camera such as:
esp32_camera_web_server with snapshot moduleesp32_camera_web_server with stream module using ffmpegffmpegLinux operating system. Debian based preferred, for example Raspberry Pi OS Lite if you run Raspberry Pi. I use also laptop with Ubuntu 22.04, but I believe with minor tweaks it should work on most distributions (mainly package names are different).
Below list uses Debian package names.
"},{"location":"requirements/#generic-system-packages","title":"Generic system packages","text":"bash 5.x (what year is it?)git (just to install scripts from this repo)curliputils-pinguuid-runtime to make generation of camera fingerprint easierv4l-utils - to detect camera capabilitieslibcamera0 - for Rpi CSI cameraslibraspberrypi-bin or rpicam-apps-lite for Rpi CSI cameras (should be already installed on Rpi OS)fswebcam - for generic USB camerasffmpeg - for custom commands for capturing remote streamsYou can run the app as container.
Multi-platform images are available at quay.io/kaszpir/prusa-connect-script.
Currently available platforms:
Install docker on Debian.
Optional - you may want to make sure current user is in docker group so it is possible to run containers without using sudo:
sudo usermod -a -G docker $(whoami)\nlogout and login again, or reboot Raspberry Pi.
"},{"location":"service.docker/#preparation-of-env-files-for-docker-command","title":"Preparation of env files for docker command","text":"Notice - you may not have to do it if you use docker-compose (I think...).
If you use docker command directly you need to edit env files and remove quotation marks from the files (this is a limitation of the Docker)
For example:
CAMERA_COMMAND_EXTRA_PARAMS=\"--immediate --nopreview --thumb none -o\"\nbecomes
CAMERA_COMMAND_EXTRA_PARAMS=--immediate --nopreview --thumb none -o\nWe assume that .csi is a env file with example variables after edit, it is possible to run below command and have screenshots sent to the Prusa Connect.
docker run --env-file .csi -v /run/udev:/run/udev:ro -v /dev/:/dev/ --device /dev:/dev --read-only quay.io/kaszpir/prusa-connect-script:03c4886\nIf you use remote camera you can make command even shorter:
docker run --env-file .esp32 --read-only quay.io/kaszpir/prusa-connect-script:03c4886\ndocker run --env-file .docker-csi --device /dev:/dev -v /dev/:/dev/ -v /run/udev:/run/udev:ro -it quay.io/kaszpir/prusa-connect-script:03c4886-arm64\n\ndocker run --env-file .docker-esphome-snapshot --read-only quay.io/kaszpir/prusa-connect-script:03c4886-amd64\ndocker run --env-file .docker-video0 --device /dev:/dev -v /dev/:/dev/ -v /run/udev:/run/udev:ro -it quay.io/kaszpir/prusa-connect-script:03c4886\nCreate env file per camera and run each container separately.
"},{"location":"service.docker/#docker-compose","title":"docker-compose","text":"Instead of running single command per container, you can manage them using docker-compose. Example docker-compose.yaml contains some examples. Some sections are commented out, though.
Notice they still require proper env files to work, for example copy usb.dist as .usb, edit its parameters and run docker-compose up
Notice that you may need to change remote cameras addresses from hostnames to IP addresses.
Another notice that sharing /dev/ or /dev/shm across different containers with different architectures may be problematic.
Depending on the distro there are various options to configure scripts as service.
Other - not implemented, do it on your own.
"},{"location":"service.systemd/","title":"Install script as systemd service","text":"Depending on the distro there are various options to configure scripts as service. On newer distros Raspberry Pi runs systemd, we will use that.
cd /home/pi/src/prusa-connect-camera-script\nsudo cp -f prusa-connect-camera@.service /etc/systemd/system/prusa-connect-camera@.service\nsudo systemctl daemon-reload\nAssuming that /home/pi/src/prusa-connect-camera-script/.env file was created in previous steps, we use that .env file as example camera config.
Notice there is no dot before env in the commands below!
sudo systemctl enable prusa-connect-camera@env.service\nsudo systemctl start prusa-connect-camera@env.service\nsudo systemctl status prusa-connect-camera@env.service\nAbove commands will enable given service on device restart (reboot), start the service and show current status.
"},{"location":"service.systemd/#configure-multiple-cameras","title":"Configure multiple cameras","text":"This project allows spawning multiple systemd units. The suffix after @ defines what env file to load from given path. For example if you set unit file name to prusa-connect-camera@csi.service then systemd will load env vars from the file under path /home/pi/src/prusa-connect-camera-script/.csi
So in short:
csi.dist as .csi and edit itprusa-connect-camera@.service as prusa-connect-camera@csi.servicecd /home/pi/src/prusa-connect-camera-script/\ncp csi.dist .csi\n# edit .csi and set custom command params, token and fingerprint etc...\nsudo systemctl enable prusa-connect-camera@csi.service\nsudo systemctl start prusa-connect-camera@csi.service\nsudo systemctl status prusa-connect-camera@csi.service\nFor another camera, let say for another camera attached over USB
cd /home/pi/src/prusa-connect-camera-script/\ncp usb.dist .usb1\n# edit .usb1 and set device, token and fingerprint etc...\nsudo systemctl enable prusa-connect-camera@usb1.service\nsudo systemctl start prusa-connect-camera@usb1.service\nsudo systemctl status prusa-connect-camera@usb1.service\nFor esphome camera, for static images:
cd /home/pi/src/prusa-connect-camera-script/\ncp esphome-snapshot.dist .esphome1\n# edit .esphome1 and set device, token and fingerprint etc...\nsudo systemctl enable prusa-connect-camera@esphome1.service\nsudo systemctl start prusa-connect-camera@esphome1.service\nsudo systemctl status prusa-connect-camera@esphome1.service\nI hope you get the idea...
"},{"location":"service.systemd/#uninstall-systemd-service","title":"Uninstall systemd service","text":"Just run two commands per camera (where csi is a camera config):
sudo systemctl stop prusa-connect-camera@csi.service\nsudo systemctl disable prusa-connect-camera@csi.service\nAfter removing all cameras remove systemd service definition and reload daemon:
sudo rm -f /etc/systemd/system/prusa-connect-camera@.service\nsudo systemctl daemon-reload\nUse mediamtx on another Raspberry Pi to create RTSP camera stream for test.
Assuming you run mediamtx with Raspberry Pi CSI camera and that raspberry-pi is the hostname of your device and that you expose two cams:
so your mediamtx.yml has config fragment such as:
paths:\n cam:\n source: rpiCamera\n\n endoscope:\n runOnInit: ffmpeg -f v4l2 -pix_fmt mjpeg -video_size 1280x960 -framerate 30 -i /dev/video1 -c:v libx264 -preset ultrafast -b:v 6000k -f rtsp rtsp://localhost:$RTSP_PORT/$MTX_PATH\n runOnInitRestart: yes\nStart mediamtx server:
./mediamtx\nThis should allow us to reach two streams, replace rpi-address with the name of your Raspberry Pi hostname or IP address. The ports are default for mediamtx.
ffplay rtsp://rpi-address:8554/cam\nffplay rtsp://rpi-address:8554/endoscope\nOr you could watch it via web browser under endpoints such as
http://rpi-address:8889/cam\nhttp://rpi-address:8889/endoscope\nRaspberry Pi Zero 2 + Logitech C920, thanks to user [&] undso.io for working example.
Allows to have a camera live stream and prusa camera script to use that stream as source of the images to send to PrusaConnect.
mediamtx config fragment
paths:\n cam:\n runOnInit: ffmpeg -f v4l2 -i /dev/video0 -pix_fmt yuv420p -video_size 1920x1080 -framerate 30 -preset ultrafast -c:v libx264 -b:v 6000k -f rtsp rtsp://localhost:$RTSP_PORT/$MTX_PATH\n runOnInitRestart: yes\nenv file for prusa connect script, remember to replace [rpizero-ip] with device address (or try 127.0.0.1 or 0.0.0.0 if script runs on the same host where mediamtx runs)
PRINTER_ADDRESS=...\nPRUSA_CONNECT_CAMERA_TOKEN=...\nPRUSA_CONNECT_CAMERA_FINGERPRINT=...\nCAMERA_DEVICE=/dev/null\nCAMERA_COMMAND=ffmpeg\nCAMERA_COMMAND_EXTRA_PARAMS=\"-loglevel error -y -rtsp_transport udp -i 'rtsp://[rpizero-ip]:8554/cam' -f image2 -vframes 1 -pix_fmt yuv420p \"\n.env is the camera config we defined earlierset -o allexport; source .env; set +o allexport\n./prusa-connect-camera.sh\nAbove commands will load env vars and will start the script. In the beginning script shows some commands that will be executed, for example command to fetch the image from camera, example log line:
Camera capture command: fswebcam -d /dev/video0 --resolution 640x480 --no-banner /dev/shm/camera_87299de9-ea57-45be-b6ea-4d388a52c954.jpg\nso you should run:
fswebcam -d /dev/video0 --resolution 640x480 --no-banner /dev/shm/camera_87299de9-ea57-45be-b6ea-4d388a52c954.jpg\nand get the outputs from the command, and also it should write an image.
Check for errors, if any, if everything is ok you should see a lot of 204 every 10s.
If not, see troubleshooting, copy logs and raise an issue on GitHub.
"},{"location":"troubleshooting/","title":"Troubleshooting","text":"Things to check if it does not work.
"},{"location":"troubleshooting/#general","title":"General","text":"some USB cameras require physical unplung/plug-in AFTER the Rasberry Pi was turned on, there is no fix for this yet, see raspberrypi/linux/issues/6255 for such example
check /dev/shm/camera_*.stdout and /dev/shm/camera_*.stderr files for more details - if they still that 'everything is okay' then probably you have issues with permissions when running script for the second time (see below)
if you use feature to ping the printer then ensure printer is up and running and responds to ping, or just disable the feature (set PRINTER_ADDRESS=\"\" or to PRINTER_ADDRESS=127.0.0.1)
check if the camera actually works - check cables if they are not damaged, if the cables are properly plugged, if the camera connects to the network...
check if the camera supports passed parameters such as resolution and codec, especially after replacing the camera - see tuning how to use v4l2-ctl to see available camera options.
check if any other app is not accessing the camera - especially local cameras are locked by another processes.
If another application is accessing camera then unfortunately only one app can access the camera and you must decide which app to run.
This means if you have something like Klipper/Obico/PrusaLink/motioneye/frigate (and so on) accessing the directly attached device to the Raspberry Pi then it will not work.
In such case you can try to find the process using fuser package, assuming /dev/video0 is your camera:
sudo apt install -y psmisc\nfuser /dev/video0\nSee StackOverflow for more details.
In general you could create a loopback camera device but this is quite a lot of work to do.
check IP/domain names for remote camera - try that you can access camera over IP address, otherwise you have a DNS issues.
file permissions - check files under /dev/shm/camera* and /dev/video0
ls -la /dev/shm/camera* /dev/video*\nand compare them with the current user executing the script or the user that is running docker (see below) or systemd service (see section below).
The quickest fix is just to delete files in /dev/shm/camera_* to fix only specific permission issues:
sudo systemctl stop prusa-connect-camera@env.service\nsudo rm -f /dev/shm/camera_*\nsudo systemctl start prusa-connect-camera@env.service\nand see if the issue is resolved.
If you still have issues due to accessing /dev/video* then ensure the user is added to video group.
dockerized script - ensure you restart the pi after adding docker, check user permissions to the mounted files and devices (unfortunately this can get very messy with direct access to the devices and files on the host)
check IP/domain names for remote camera - ensure that you can access camera over IP address (or fully qualified domain name), because .local or .lan domains are not resolved. Another option is to reconfigure docker to use proper local DNS servers and not generic 8.8.8.8.
You can also try to run container with --add-host or extra_hosts in docker-compose.
Another option is to run container with --network=\"host\" or network_mode: \"host\" in docker-compose.
"},{"location":"troubleshooting/#systemd-troubleshooting","title":"Systemd troubleshooting","text":""},{"location":"troubleshooting/#get-systemd-logs","title":"Get systemd logs","text":"If the script runs locally but service is not running then you can get the logs like below, ensure to replace env with the name your camera is using:
sudo systemctl stop prusa-connect-camera@env.service\nsudo journalctl -f -u prusa-connect-camera\nand keep it open
sudo systemctl start prusa-connect-camera@env.service\nsleep 10\nsudo systemctl stop prusa-connect-camera@env.service\nget back to the terminal with running journalctl and see the logs and look carefully at the errors described there
copy the output from starting to the another starting command and paste on GitHub
nano editorsudo nano /etc/systemd/system/prusa-connect-camera@.service\nand replace User=pi and Group=pi with the current user and group, then reload systemd and start service again
sudo systemctl daemon-reload\nsudo systemctl start prusa-connect-camera@env.service\nThis way it will use your user account to access camera device and write files.
"}]} \ No newline at end of file diff --git a/service.docker/index.html b/service.docker/index.html new file mode 100644 index 0000000..5a0b6af --- /dev/null +++ b/service.docker/index.html @@ -0,0 +1,1137 @@ + + + + + + + + + + + + + + + + + + + + + + + +You can run the app as container.
+Multi-platform images are available at quay.io/kaszpir/prusa-connect-script.
+Currently available platforms:
+Install docker on Debian.
+Optional - you may want to make sure current user is in docker group so it is possible
+to run containers without using sudo:
logout and login again, or reboot Raspberry Pi.
+Notice - you may not have to do it if you use docker-compose (I think...).
+If you use docker command directly you need to edit env files
+and remove quotation marks from the files (this is a limitation of the Docker)
For example:
+ +becomes
+ +We assume that .csi is a env file with example variables after edit, it is
+possible to run below command and have screenshots sent to the Prusa Connect.
docker run --env-file .csi -v /run/udev:/run/udev:ro -v /dev/:/dev/ --device /dev:/dev --read-only quay.io/kaszpir/prusa-connect-script:03c4886
+If you use remote camera you can make command even shorter:
+ +docker run --env-file .docker-csi --device /dev:/dev -v /dev/:/dev/ -v /run/udev:/run/udev:ro -it quay.io/kaszpir/prusa-connect-script:03c4886-arm64
+
+docker run --env-file .docker-esphome-snapshot --read-only quay.io/kaszpir/prusa-connect-script:03c4886-amd64
+docker run --env-file .docker-video0 --device /dev:/dev -v /dev/:/dev/ -v /run/udev:/run/udev:ro -it quay.io/kaszpir/prusa-connect-script:03c4886
+Create env file per camera and run each container separately.
+Instead of running single command per container, you can manage them using
+docker-compose. Example docker-compose.yaml contains some examples.
+Some sections are commented out, though.
Notice they still require proper env files to work, for example
+copy usb.dist as .usb, edit its parameters and run docker-compose up
Notice that you may need to change remote cameras addresses from hostnames +to IP addresses.
+Another notice that sharing /dev/ or /dev/shm across different containers
+with different architectures may be problematic.
Depending on the distro there are various options to configure scripts as service. +On newer distros Raspberry Pi runs systemd, we will use that.
+cd /home/pi/src/prusa-connect-camera-script
+sudo cp -f prusa-connect-camera@.service /etc/systemd/system/prusa-connect-camera@.service
+sudo systemctl daemon-reload
+Assuming that /home/pi/src/prusa-connect-camera-script/.env file was created in
+previous steps, we use that .env file as example camera config.
Notice there is no dot before env in the commands below!
sudo systemctl enable prusa-connect-camera@env.service
+sudo systemctl start prusa-connect-camera@env.service
+sudo systemctl status prusa-connect-camera@env.service
+Above commands will enable given service on device restart (reboot), +start the service and show current status.
+This project allows spawning multiple systemd units.
+The suffix after @ defines what env file to load from given path.
+For example if you set unit file name to prusa-connect-camera@csi.service
+then systemd will load env vars from the file under path
+/home/pi/src/prusa-connect-camera-script/.csi
So in short:
+csi.dist as .csi and edit itprusa-connect-camera@.service as prusa-connect-camera@csi.servicecd /home/pi/src/prusa-connect-camera-script/
+cp csi.dist .csi
+# edit .csi and set custom command params, token and fingerprint etc...
+sudo systemctl enable prusa-connect-camera@csi.service
+sudo systemctl start prusa-connect-camera@csi.service
+sudo systemctl status prusa-connect-camera@csi.service
+For another camera, let say for another camera attached over USB
+cd /home/pi/src/prusa-connect-camera-script/
+cp usb.dist .usb1
+# edit .usb1 and set device, token and fingerprint etc...
+sudo systemctl enable prusa-connect-camera@usb1.service
+sudo systemctl start prusa-connect-camera@usb1.service
+sudo systemctl status prusa-connect-camera@usb1.service
+For esphome camera, for static images:
+cd /home/pi/src/prusa-connect-camera-script/
+cp esphome-snapshot.dist .esphome1
+# edit .esphome1 and set device, token and fingerprint etc...
+sudo systemctl enable prusa-connect-camera@esphome1.service
+sudo systemctl start prusa-connect-camera@esphome1.service
+sudo systemctl status prusa-connect-camera@esphome1.service
+I hope you get the idea...
+Just run two commands per camera (where csi is a camera config):
sudo systemctl stop prusa-connect-camera@csi.service
+sudo systemctl disable prusa-connect-camera@csi.service
+After removing all cameras remove systemd service definition and reload daemon:
+ + +Depending on the distro there are various options to configure scripts as service.
+ +Other - not implemented, do it on your own.
+ +