dralexliu / octoprint-pinozcam

AI model running on RPi for failure detection

OctoPrint-PiNozCam

Introduction

Welcome to the era of edge computing with free failure detection performed directly on your Pi

Device (50% of All Cores for AI)	Inference Speed (images / minute)
Raspberry Pi 5	45
Raspberry Pi 4	9
Raspberry Pi 3B	5
PC with Intel i5 10600	260
OrangePi Zero 2/3	9
Raspberry Pi Zero 2 W	3

_{*The inference speed tests were conducted under the circumstance that 50% of the device's cores were allocated for AI processing, while the remaining 50% of the cores were dedicated to OctoPrint and printing processes.}

Unlock advanced 3D printing monitoring with PiNozCam, your go-to solution for AI-powered surveillance — all without any subscription or registration. PiNozCam brings cutting-edge computing to your Raspberry Pi or any old PC/single board computer, ensuring privacy and providing instant failure alerts via Telegram/Discord.

Features include:

• Fast Inference on Pi Arm CPU/Local Device, 24/7 AI service
• Instant Telegram/Discord Error Notifications
• Privacy-first approach with RAM-only data storage
• No email register/sign up/subscription/Cloud/Ads/Payment
• Auto Pause/Stop Functionality

Support RPi OS platform (Don’t know❓):

OS platform	Buster	Bullseye	Bookworm
arm64 (x64)	✅	✅	✅
armhf (x32)	✅	✅	✅

⚠️ This plugin supports the OctoPi image . However, I am still working on this plugin on Octo4a and it may be supported in the future versions.

RPi(Boardcom)	Intel/AMD CPU	AllWinner	RockChip	RAM
✅	✅	✅	✅	>=512MB

Plugin Setup

Install via the bundled Plugin Manager or manually using this URL:

https://github.com/DrAlexLiu/OctoPrint-PiNozCam/archive/master.zip

Required Hardware Setup

Endoscope Camera

Compatible with most market-available endoscope cameras.

• Nozzle camera kits: StealthBurner, 3Do, etc.
• Build yours with cameras from Aliexpress or Amazon or Temu.

Ensure your camera:

• 30Hz frame rate, 16:9, >=480P❓
• Built-in LED backlighting.
• Positioned 5-10 cm from the nozzle.

⚠️ Cleaning the camera lens before EACH print is highly recommended for dust removal.

Endoscope camera Bracket

Search and print a nozzle camera bracket for your camera model.

The setup would be like this:

Cellphone

To enable notifications, enter your Telegram bot token and chat ID or Discord Webhook url. Upon configuration and clicking "Save". A welcome message confirms successful setup. An example failure notification will be sent like this:

Software Configuration

Navigate to the PiNozCam tab:

The screenshot:

Key Parameters:

• Action: Specifies the action PiNozCam should take when a print failure is detected (e.g., notify only, pause print, stop print). Detected failures are displayed in the video stream for 5 seconds, allowing for immediate visual verification.
• Image Sensitivity: Adjust the sensitivity to ensure accurate detection of print failures. Set the threshold to balance between premature stopping for minor issues and delaying action for significant errors. A starting value of 0.04 or 4% is recommended for optimal balance.
• Failure Scores Threshold: Define the confidence level at which an anomaly is considered a print failure. This setting helps in reducing false alarms by setting a minimum probability threshold for errors, ensuring that only genuine failures prompt action.
• Max Failure Count: Specify the number of detections required in Failure Consider Time before PiNozCam takes the configured action. A value above 1 is recommended to avoid false positives.
• Failure Consider Time (s): Implement a time buffer to focus on recent failures, ignoring older detections that may no longer be relevant. This dynamic consideration helps in adapting to the current state of the print.
• CPU Speed Control: Offers options for running the CPU at half or full speed. Half speed is recommended.

Initially, stick with the default settings and adjust them gradually to fine-tune performance.

Support

For further discussion and support, please join our Discord channel.

Support my work

I created this plugin in my spare time, so if you have enjoyed using it then please support it’s development!

octoprint-pinozcam's Issues

Discord Link is Invalid

The Discord link is invalid.

PiNozCam Camera View Unchangable

My Camera is mounted "upside-down" but is of course adjustable in Octoprint camera control.

How can this be achieved with PiNozCam and does this matter?

Install Not Getting Requirements

There have been multiple reports on Discord and in the forum of people installing this plugin and it not loading properly because it's missing dependencies. Here's one example.

https://community.octoprint.org/t/pinozcam-plugin-unable-to-show-up/57331

I suspect it's from your conditional setup_requires possibly?

[Feature Request] Additional notification options.

So far, this thing is fantastic. thank you!!!

Would it be possible to add additional notification options other than telegram, even just a generic url/data packet to send to an API?

Installation error on octo4a

hey, when installing this plugin I get this error:
ERROR: Could not find a version that satisfies the requirement onnxruntime>=1.15.0 (from octoprint-pinozcam) (from versions: none)
ERROR: No matching distribution found for onnxruntime>=1.15.0

which results in the message: "Installed: Unknown"

I have python 3.10 installed, and from what I can tell, onnx is inly supported up to 3.9

I am using a Samusng Galaxy Note 8 as my device using octo4a

Feature request: max failure messages per print

When a failure is detected I get Telegram messages until the print is stopped. Then my phone gets crazy :-).

I would like to request a limit input field for failure messages per print. E. g. setting this to "5" would only send max. 5 messages within one print job.

There should be a option to switch the limit off. Could be reached by "0" as input value.

assumes octopi 1.0.0

This line appears to assume bullseye image, but there are also still stretch and bookworm releases out in the wild.

https://github.com/DrAlexLiu/OctoPrint-PiNozCam/blob/1c9519ced05f8e41434b57615cf8a81be97dcff3/setup.py#L40C5-L40C159

ERROR: onnxruntime-1.15.0-cp39-cp39-linux_armv7l.whl is not a supported wheel on this platform.

Installation Error on pi4

First of thanks for making this! Very excited to try it out. I generally don't love opening issues but I did google as much as I could and can't really find anything about this error. Also python is very much not my wheel house or I would try and dig in further.

Anyway I am trying to install it via the plugin manager and getting the following error on a pi4

`Installing plugin "OctoPrint-PiNozCam" from https://github.com/DrAlexLiu/OctoPrint-PiNozCam/archive/master.zip...
/home/pi/oprint/bin/python3 -m pip --disable-pip-version-check install file:///tmp/tmpt9vqio7l/OctoPrint-PiNozCam-master.zip --no-cache-dir

Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple
Processing /tmp/tmpt9vqio7l/OctoPrint-PiNozCam-master.zip

ERROR: onnxruntime-1.16.0-cp39-cp39-linux_armv7l.whl is not a supported wheel on this platform.

Error!
Could not parse output from pip, see plugin_pluginmanager_console.log for generated output`

100% CPU on Pi 4 8Gb leads to bulbs while printing

I installed the plugin on my Pi 4 8 Gb running

Linux octopi 6.1.21-v7l+ #1642 SMP Mon Apr  3 17:22:30 BST 2023 armv7l
OctoPrint version : 1.9.3
OctoPi version    : 1.0.0
PiNozCam (1.0.0)

I used the recomended settings for Pi 4 but the system gets to 100% CPU each time the snapshot is taken. This leads to pauses in the print head´s movement. At each pause the suface of the print gets a bulb.

When I disable the plugin the surface of the print is smooth again.

AI Detects Print Head Grill As an Issue

The AI consistently detects the Print Head Grill as an issue. Please see image.