License Plate Detection Demo Using Raspberry Pi Camera

No more parking tickets?
Product Links (some are affiliate links)
- Raspberry Pi 5 👉 amzn.to/3SzmYiT
- Raspberry Pi Camera 👉 amzn.to/3u9ewh8
This tutorial will demonstrate how ip cameras or Raspberry Pi with camera module can be used to both detect and transcribe license plates in video feeds. This particular implementation uses a license plate recognition software package called Open ALPR.
OpenALPR is a NVIDIA software company that develops license plate and vehicle recognition products. The OpenALPR software achieves state-of-the-art accuracy through the use of deep learning. OpenALPR products are used in a wide variety of applications, such as surveillance, parking enforcement, data entry, and supply chain automation. The company is based in Boston, MA and supports customers worldwide.
ALPR, ANPR, LPR, License Plate Recognition, Open Source, Free, License Plate, Number Plate
Outline
Intro 0:00 - 1:02
Parking Garage Prototype 1:02 - 1:55
Open ALPR 1:55 - 2:28
Tech Stack 2:28 - 3:26
Configure RPI 3:26 - 4:00
Add Camera Module 4:00 - 4:34
Format MicroSD Card 4:34 - 5:00
Install Imager 5:00 - 5:22
Advanced Options 5:22 - 6:23
Boot RPi 6:23 - 7:30
Connect to RPi with SSH 7:30 - 14:10
Create AutoML Model 14:10 - 41:33
GitHub Repo
github.com/aioverlords/Licens...
#computervision #raspberrypi #openalpr
Open ALPR
www.openalpr.com/
www.tensorflow.org/js
tesseract.projectnaptha.com/
www.raspberrypi.com/software/

[@AtactHD]

I can't believe I just watched the full hour of this. Definitely one of the coolest videos I've ever seen. Thank you!

[@johnbubu8310]

Thank you for your tutorial. It gives a good overview of how ai models work 😊

[@DaisyHollowBooks]

This is really interesting. I had no idea this kind of thing is possible with Raspberry Pi.

[@AndersJackson]

Great video on a interesting topic. thanks. Some comments: I would set the name when flashing the SD card with Imager program. Then there will not be a conflicting name when you connect with ssh. And IF you get a conflict, you can also use the command line to remove the conflicting name from ssh configuration, if you have. I never do "sudo su -" I just use "sudo commad" when I need root access. I prefers Emacs instead of vim. I usually edit files over ssh by using Emacs on my machine. I find that easier then to install Emacs or vim on the target machine.

[@JohnDoe-id9hi]

Thank you, thank you so much for making this video.

[@trezero]

Great video. Any significant updates with using Vertex AI since it seems to be replacing the older Vision AI?

[@kerron_]

Will try this

[@mikkolempinen2717]

Does this learn from every new camera photo or do you have to manually make a new run with all the images?

[@tacom6]

What if all I need is an endpoint that reads numbers off of a camera feed (source could be another monitor with some metrics etc etc). Is there a way to accomplish that without using Web APIs, ie completely offline.

[@SH0RELESS]

Thx for effort

[@meghanarprakash7862]

Great video!!! It helped a lot. I just wanted to know the camera specification

[@jaygeeantonio]

Can you write a sql injection on paper and use the ocr in speed cams to do anything

[@yulianad.sabusalu8592]

I really like your video, it's really good. I'm working on my thesis. I am very grateful if you can help me.🙏

[@thiruvengadam96]

Can you compare Raspberry Pi with Google Coral Dev Board for similar use case?

[@SmarTechZ]

Connecting a Coral TPU to a Raspberry Pi will enhance speed, but is it necessary to convert the model to TensorFlow Lite?

[@vikashkumar-cr7ee]

Dear tutor Greetings! I have gone through your KZbin tutorial lector and found it really informative. I am requesting you make a Tutorial integrating IoT Raspberry Pi and computer vision. It would be highly helpful for students like us who want to develop a project.

[@user-ed4yp6eq5k]

Question. Im using roboflow on my surface pro and steam deck my model is getting 5fps with this be faster? I need at least 30fps

[@kilianlanger3668]

Danke!

[@toober6789]

Do you think it'd be possible to match the license plates against a database of stolen cars and then flash a dynamic sign to drivers to let them know they are driving a stolen car? (This could be a useful deterrent against crimes committed in stolen cars!)

[@alwaha4283]

Very informative video tnx😮

[@muhammadm1237]

Which os version you are running on the pi? Year and bit type (32-64bit)

[@jirayuvijjakajohn295]

51:26 what is it that you are holding in your left hand

[@mintheartist8890]

Data Slayer. At 8:17 when you clear out the host file. For me when I try this, it doesn't let me edit it at all. Is there a solution to this?

[@ma.sofhiahorteza6164]

hi there data slayer been following your method i tried to import the csv file into the datasheet it seems it does not accept it

[@samyakgauri2749]

Does this Work with Speedy Moving Vehicles?

[@Anonymous-qj5no]

Hi can i use it with a usb webcam

[@arhaanar1577]

Can anyone please provide me with the link to buy the camera module or just tell me the camera model i really need it .... project deadline is approaching please help me out ....

[@lunaestella1384]

can i integrate it with using only a webcamera?

[@jordancreglia5754]

Thoughts on how to set this up with an existing PoE camera and DVR?

[@RicardoRibeiro-dorcari]

If you can get an rtsp stream of the video feed you can make it work

[@PrivateUsername]

Comment for the algo: man, I can't believe you are able to get any work done at all with that horrible font in the Terminal.

[@aymaneelfahsi8048]

Hello sir, will this work with raspberrypi4 with a pi camera module 2 (i’m planning onusing opencv-python)

[@abc123fhdi]

can you do this with a dash cam.

[@caryllcarino-lg4fi]

is it applicable on windows?

[@KettlebellGunsmith]

Anyone else notice it OCR'ed the "GDD Z55" license plate incorrectly as "GDD 255"?

[@blueraventy]

Is it okay to use usb drive instead of sd card?

[@LuisCarlosMalo]

Thanks for sharing!! Do you know some 100% local alternative without having to send the pictures to the internet? A raspberry alone may not have enough power to analyze the number plates by itself, but what about with the help of a Google Coral USB stick or similar? Thanks in advance!

[@DataSlayerMedia]

It has enough power I just couldn't fine any good local OCR that worked well. There are probable some out there tho.

[@LuisCarlosMalo]

@@DataSlayerMedia thanks 🙏

[@arsenkabey2684]

Hello Dear, please a need to extract plate from movies

[@bccicone]

Great video! Could you list the commands and the websites you are using. Having a hard time seeing the commands. Much appreciated!! Not seeing them in the text below the video.

[@bccicone]

Just FYI.. I have everything completed up to the start of google cloud and forward.

[@virugupta707]

This model will be effective when vehicle is moving

[@harshadade2849]

how's if we are not using led display

[@a.t.mfazlarabbi2528]

What's the ram size of this raspberry pi 4 module B ?

[@anuragparcha4483]

4 GB I believe

[@_HI-tt2qp]

@49:49 I haver error Cannot find module '@google-cloud/vision' .... How can I solved it?

[@mr.manfresh4244]

did you manage to solve it?

[@barachielling8245]

same

[@LordJupiter7]

I'm also faced with this problem. A reply or a fix to the git repo would be appreciated.

[@netq254]

The ending is confusing me, you say that Tesseract is inaccurate and to use Google Vision for the extraction of the license plate string, but then you don't explain how to use it or set it up, what am I missing?

[@nauffaliqbal7921]

yeah i feel you, do u by any chance know how to use/set it up now?ur replies mean so much to me

[@netq254]

@@nauffaliqbal7921 I went a completely different route. Easy OCR for python is the most effective tool I found. The key is: High resolution images of the plate, (the more zoomed in and clearer the image, the more accurate the results) and you may need the necessary pre-processing steps on the image before attempting to run OCR on it. Google common pre-processing tips for OCR or ask chatGPT (grayscale, thresholding etc).

[@caplanrobert8288]

great tutorial! but I have to say it, your terminal character spacing is kinda weird :D

[@coucal]

🙂 26:20 is Sanjay Dutt's ROLLS ROYCE PHANTOM. He is a India film actor.

[@DataSlayerMedia]

Interesting

[@shubhamsahamate6188]

fyi it's a RR Ghost, not a phantom

[@mahmoudhussien5522]

Hello where is a gcp-key.json script?

[@mr.manfresh4244]

did you find it?

[@samppaturunen9371]

I replicated what you did and eventually got it working since the Google Cloud has changed a bit and, you did not cover all details like creating credentials and such. But thanks!

[@stark_7114]

Brother can you help me through this code running issue. where is gcp-key.json script? how to run the script?

[@nauffaliqbal7921]

did u know how to set up the google vision since he doesnt show it

[@johncarpenter224]

Business format waiting

[@MYance-ut3ud]

How can i get code?

[@DataSlayerMedia]

github.com/aioverlords/License-Plate-Detection

[@NFlight]

Funny... this video is only a few months old and the areas in the video on Google look way different now.

[@weirdsciencetv4999]

Why not also train the model on the edge? I am pretty aggravated by cloud services

[@ericmbusa]

Both of your videos are giving me js errors .. possibly related to the version? What version were you using in this video? (function (exports, require, module, __filename, __dirname) { const { exec } = require('child_process'); SyntaxError: Unexpected token {

[@evanellis4097]

You didn't close the bracket, should be (function (exports, require, module, __filename, __dirname) { const { exec } = require('child_process')};

[@Andres-Estrella]

I can't believe a raspberry pi can run these neural networks

[@rahulaggarwal3977]

What?

[@caribbeanchild]

Cold coffee. Yuck.

[@MohamadsGarage]

bro, are you ok? you sounds sleepy asf

[@meant4com]

Well using cloud vision endpoint is contradicting you living on the edge