REAL TIME Number Plate Recognition with Python and AWS | Object detection and tracking | Yolov8

Code: github.com/computervisioneng/...
🎬 Timestamps ⏱️
0:00 Intro
0:25 Start
1:21 High level description / Super detailed description
2:45 Amazon Kinesis Video Streams
4:22 Setup producer
11:37 Setup consumer #1: Object detection and tracking
34:50 Setup consumer #2: Visualization
48:35 Mind the costs
48:59 Dive into the details
49:15 Outro
🌍 Community 👥
Join our Discord server: / discord
Support me on Patreon: / computervisionengineer
#python #objectdetection #objecttracking #yolov8 #aws #sort

[@ComputerVisionEngineer]

FULL COURSE is available in my Patreon to all the Computer Vision Experts: www.patreon.com/posts/real-time-number-93087151 😃🙌

[@dimplekhuman9147]

Tysm for this awesome tutorial. btw your accent is very soothing...

[@ComputerVisionEngineer]

Thank you! Glad you enjoyed it! 🙌

[@Muhammad_Adams]

Thank you for the hard work......keep it up

[@ComputerVisionEngineer]

Thank you! 😃🙌

[@gallorate]

Great Felipao! you`re the best

[@ComputerVisionEngineer]

Thank you for your support! 😃🙌

[@melissaakli4645]

so intersting! thank you !

[@codation]

Thanks for this great tutorial. If I have a real-time traffic video feed coming in from a CCTV, how can I stream it on AWS?

[@tello9504]

Como quedaría la arquitectura si los datos los disparó desde una cámara con jetson? Que patrón me recomiendas?

[@nikhilkarnik6961]

Can you point me to the code/library which is responsible for the license plate detection?

[@readbhagwatgeeta3810]

Hello! I am not able to fund your lane change detection video which you had uploaded. Can you please share me the link of that video 🥺

[@nguyenthehoang9148]

Great, thank you for the hard work! What foundational knowledge would you recommend before entering this full content to get the most out of it?

[@ComputerVisionEngineer]

Mmm I think Python and some familiarity with the AWS environment are the only prerequisites. 🙌

[@danielhurtado3493]

@ComputerVisionEngineer why did you use EC2 instead of Sagemaker? is it possible to do it in sagemaker?

[@ComputerVisionEngineer]

Hi, do you mean if it is possible to do any part of this process in Sagemaker instead of EC2? Perhaps training the object detector and creating an API to make inferences with it can be done in Sagemaker. Setting up the producer and consumer can't be done in Sagemaker.

[@satishchintal]

Hi @ComputerVisionEngineer is there any way we can have the automatic deployment without using the manual way of triggering the producer, consumer and local. where we can have one script to automate the process

[@ComputerVisionEngineer]

It may be possible using a CloudFormation template, I would need to look into it.

[@AlbertoRobles72]

Great video! Have much does the infrastructure costs?! Have you done an estimate? Is it the best option?

[@ComputerVisionEngineer]

Thank you! I haven't done an estimate, but I may prepare it later on and upload it. It would be nice to have an estimate of cost / minute of video, or similar. 😃💪

[@rafaeldalpra]

I would prefer something more local than in the cloud, but the project is still interesting.

[@persevere1052]

Awesome vid! I am training a YOLO model to detect and segment particular mechanical devices. The accuracy never goes up more than 85-90% while training. Could you please suggest a few strategies to improve the situation? Thanks!

[@ComputerVisionEngineer]

It is probably due to the data, collect more images and make sure the labels are correct. 🙌

[@persevere1052]

@@ComputerVisionEngineer There are around 20 classes. How many images do you suggest to collect?

[@ComputerVisionEngineer]

@@persevere1052 A few hundreds per class at least, also make sure the labels are correct.

[@persevere1052]

@@ComputerVisionEngineer Thank you so much for your inputs! Will bother you with a last question though, is it fine to group different objects i.e. different shapes under the same label (as we've to do manual annotations) or would that hit the mAP?

[@vamsikatari3112]

Awesome Video . I am trying to do automatic annotation for large datasets to reduce the time of manual annotations, Can you please make a video or suggest me any idea regarding that.

[@ComputerVisionEngineer]

I will try to make a video. 🙌

[@elhnston6589]

Now, we need to do this localy...

[@victorfdezbau]

Hi, thx for your work. I've a problem. I've tried to follow your instructions but there are missing steps in your video. I assume you copied 'aws-ft-key-pair.pem' generated by amazon on ~/.ssh folder and then ran ssh command. I made the same, but when I ran the ssh command I got the following error "Permission denied (publickey)". I put the same user name as you on the ssh comman "ssh -i ~/.ssh/aws-ft-key-pair.pem ubuntu@[instance ip]". Any piece of advice? thx in advance

[@ComputerVisionEngineer]

Hi, have you generated a key pair when launched the ec2 instance?

[@victorfdezbau]

yes, I've generated a key pair and named it like you did in the demo@@ComputerVisionEngineer

[@Ygbarakat]

@@victorfdezbau could be the permissions on the key pair you are using. It's not a networking issue at all since you wouldn't receive the permission denied error if the server was unreachable

[@jackito0506]

the t2.xlarge instance is powerful enough to perform the inference in real time?

[@ComputerVisionEngineer]

I am applying a few strategies to make the inference as close to real time as possible. 🙌

[@jackito0506]

@@ComputerVisionEngineer which strategies?

[@manoharakmanu8134]

Hi sir , i have a dout can I train multiple model and use in same program if it is possible please suggest

[@ComputerVisionEngineer]

Hi, what do you mean with 'train multiple model'?

[@manoharakmanu8134]

Like one model I traine for human detection another model for animal detection like that I trained using yolo8 after that I same frame of output can I detect and label the object in both model

[@devdiptabiswas3072]

will this work with the base free version available for aws?

[@ComputerVisionEngineer]

Some of the aws services we use in this video are not in the aws free tier.

[@lanreuzamere4994]

Thanks a thousand 🙏 . . . It's been a while though 😇

[@ComputerVisionEngineer]

Yeah it is finally ready. 😅 You are welcome! 😃🙌

[@jeremyterrible1692]

hello, what version of python are you using?

[@ComputerVisionEngineer]

In this tutorial I am using Python 3.10 if not mistaken.

[@KrithikRoshan00]

does this use yolov8 model or textract?

[@ComputerVisionEngineer]

I use both. 😃🙌

[@aburriyo]

What you think of doing it inside of a Jetson?

[@ComputerVisionEngineer]

It would be a good idea for a future project. For this one I was trying to build the entire system in AWS. 🙌

[@user-sq8qx4gh3g]

does this work on raspberry pi 4b?

[@ComputerVisionEngineer]

I haven't tested it on a Raspberry pi, if you use a Raspberry pi let me know how it goes!

[@MDAsadullahShibli]

I have a question, in projects, custom model vs pretained model , which one we should use ? @ComputerVisionEngineer

[@ComputerVisionEngineer]

It depends, if pre trained model works well for the project you are working in, you can just go ahead and use it instead of training a custom model. 🙌

[@nirangalakshan3877]

Can you do CNN/RNN tutorial 🥺

[@ComputerVisionEngineer]

I will try to 🙌

[@user-su7el9rh4y]

your project is not working on different videos

[@ComputerVisionEngineer]

Hi, what type of 'different videos' are you trying this project with?