Simple A.I. Demo with Jetson Nano
Вставка
- Опубліковано 14 лип 2024
- Get your first 10 PCBs for free at www.pcbway.com/
I'm trying out training a really simple AI machine learning model using transfer learning on the NVIDIA Jetson Nano with Jetson Inference. I used a very minimal data set of images captured and trained using scripts provided by NVIDIA. I wrote a simple script to make the robot look for high contrast markers in turn.
Check out more Jetson Projects: nvda.ws/3gAmKm9
Jetson Inference docs: github.com/dusty-nv/jetson-in...
Autonomous racer video: • DIY Autonomous Car Rac...
You can support me on Patreon or buy my Merchandise:
***************************
Patreon: / xrobots
Merchandise: teespring.com/stores/james-br...
***************************
Affiliate links - I will get some money of you use them to sign up or buy something:
***************************
Matterhackers 3D printing supplies: www.matterhackers.com?aff=7500
Music for your UA-cam videos: share.epidemicsound.com/xrobots
***************************
Other socials:
***************************
Instagram: / xrobotsuk
Facebook: / xrobotsuk
Twitter: / xrobotsuk
***************************
CAD and Code for my projects: github.com/XRobots/tracks/tre...
Thanks to 3D Fuel for the filament: www.3dfuel.com/
Huge thanks to my Patrons, without whom my standard of living would drastically decline. Like, inside out-Farm Foods bag decline. Plus a very special shoutout to Lulzbot, Inc who keep me in LulzBot 3D printers and support me via Patreon.
HARDWARE/SOFTWARE
Below you can also find a lot of the typical tools, equipment and supplies used in my projects:
Lulzbot 3D Printers: bit.ly/2Sj6nil
Lincoln Electric Welder: bit.ly/2Rqhqos
CNC Router: bit.ly/2QdsNjt
Ryobi Tools: bit.ly/2RhArcD
Axminster Micro Lathe: bit.ly/2Sj6eeN
3D Printer Filament: bit.ly/2PdcdUu
Soldering Iron: bit.ly/2DrNWDR
Vectric CNC Software: bit.ly/2zxpZqv
Why not join my community, who are mostly made up of actual geniuses. There’s a Facebook group and everything: / 287089964833488
XROBOTS
Former toy designer, current UA-cam maker and general robotics, electrical and mechanical engineer, I’m a fan of doing it yourself and innovation by trial and error. My channel is where I share some of my useful and not-so-useful inventions, designs and maker advice. Iron Man is my go-to cosplay, and 3D printing can solve most issues - broken bolts, missing parts, world hunger, you name it.
XRobots is the community around my content where you can get in touch, share tips and advice, and more build FAQs, schematics and designs are also available. - Наука та технологія
Patrons and UA-cam Channel Members can get access to all my projects up to a week early! www.patreon.com/XRobots
OpenDog doing tricks in response to hand gestures would be amazing and adorable. I can't wait for that day to come xx
That is already being planned
I was hooked on machine learning but lately i never can find to many videos giving good use cases for it but now here is the perfect video explaining it so now i can have another project to do. Keep up the good videos James.
There's a lot more coming
Great introduction to the Jetson Nano! I got one and I'm going to try building the jetbot, thanks for the pointers you gave in this video :)
You are a legend...so happy you're working with the jetson nano
Nice video like usual James, keep the good work up!
Again great job, congratulations James ...
Thanks! Really interesting to see how this stuff goes.
This guy and Alex lab are so inspiring
This project looks deceptively simple for how actually complicated the software that does the Machine Learning actually is. gg to NVidia for making the technology so accessible! I am sure you will make actually something amazing! Seeing how difficult it is to recognize a simple square, imagine how difficult it must be for an autonomous auto.
I have facial expression coming up in a couple of weeks - that was quite easy as well though
как раз посмотрел прошлую часть и вот тебе на, новая часть, спасибо за видео 👍🏻 (а также я во время просмотра прошлого видео подумал что не мешало бы добавить в этот проект камеру и вот, так и произошло)
Just what I needed thank you !
Just learned about the jetson nano like a week ago, very timely
raspberry pi is the mid price alternative, arduino the cheap alternative
Great project
Nice job👍🏽
Very brave to try and recognise a triangle when the sofa is made of triangles :D
Hi James, I really enjoyed this iteration of the build project. I am very impressed with the learning simplicity. Did you get any conflicts while roaming in the living room since your polka dot curtains contained the black circle patterns of similar size? Thanks again for this informative video.
No, it needs to see the shape on a white background, so it's pretty reliable
The music is a banger
Products like the Jetson scream code me! There's something really special in training the computer's AI that I enjoy the most.
It's funny that your room had decorations including circles squares and triangles but it only saw your markers.
beautiful =)
Printed side holes were great ! I think you have multiple printers, could you say which one you printed the parts?
Could the circle turning into a triangle be the shake of the camera? As this wasn't happening with a steady camera. I guess the refresh of the camera could be playing a part too. Nic video explaining something so complex 👍 ⭐⭐⭐⭐⭐
I want to build this! Is there STL available to print? This is so cool !
"What is my purpose" oh my god...
Great video! Are all the the electronics on the robot powered by the 24V battery? I'm working on a small project and currently using different power sources for the SBC and the motors, is it ok if i conect everything to the same power source using voltage regulators?
gesture recognition would be sick. I want to do it as well, but first i have to figure out the camera setup without one of those nvidia kits. Can it be done using a raspberry pi?
It would be cool to recognise gates positions, detecting the distance they are from the size they look on the camera, and detecting their orientation angle based on the shape that is appearing on the camera vs the original shape
Yes probably, pretty much any marker/object would do.
@@jamesbruton I thought of it like a race checkpoint and the tank could make its way through many gates
Like in the DRL autonomous race
Can you train a neural network with gestures? Like actual hand movements or just static hands?
Thanks! Really interesting to see how this stuff goes. How to Run a Python Script Automatically on Bootup / Startup in Jetson nano ?
What's the green graph kind of sheet on the desk called? Seen that in many videos
Do you plan on navigating the car with a depth map?
You will probably need stereo vision with two rpi cameras like the one you're using. There's a HAT by Arducam and Stereopi. An alternative is Intel Realsense or ZED 2.
If your circle is turning into a triangle ... you gonna have a bad time.
Hey there^^
ive seen youve done a few videos on robots with inverse kinematics, tho you dont show how you do it?
ive built myself a 18 dof hexapod with an arduino and a pololu servo controller, yet i have no idea on how to install an inverse kinematics code on there to control the servos(i dont want to use a frame by frame program tho)
ive subbed so should you decide to do a video on it ill make sure to watch it^^
nice work; keep it up :thumbsup:^^
Edit; i wanna control the robot via a rc radio, but i think using a jetson could be interesting as wellXD
The robot at 12:00: "What the fuck is that!?"
So glad python is really a thing now in IOT :-D
Please can you do a video on your workflow? How do you manage to design and print so many objects a week? Then you add in vast amounts of programming. Do you have a large team behind the camera?
when you can answer your questions start acting on them :)
No, it's just me and a Trello board
I've always wanted my own securitrons lol
interesting @ 11:40 I thought the robot is gonna take the circles in the curtain as indicators, maybe they are too small compared to your print?
The mad lad made a robotic puppy
You, the Hacksmith and Mark Rober are the smartest men on UA-cam! Keep up to good work!! Everybody that reads this: stay positive and have a great day!!!
I want to create a trash collecting robot with my nephew. Could this be the base starting kit? I know how to use object detection for seeing bottles, plastic, paper etc. Any thoughts? I appreciate your advice. Great video.
I'm going to use this to vision system to put away the dishes!
Can it be trained to recognize features on an object? How does it do with partially covered objects?
With the training you did, at what angle (off of the normal to the paper) will the ai not recognize the shape?
@@TheRainHarvester in this case the training data was pretty sparse. If he trained with more angles it would work for more angles
I want the dc motor having encoders using Jatson not with Arduino boards, how can we do it?
Is it possible some how that macbook Air M1 can be use an monitor for Jetson board ?
Please reply I am new to this field
Hello I just got myself a Jetson Nano and a Sparkfun Jetbot Kit. All is going great yet before I start up on programming I need to find a way to install a Wheel Encoder. I'm at a loss really since sparkfun does sale the wheel encoders yet do not say what is NEED to control them. I want to build a robot which will teach itself but for this I need a way for it to be able to sense when it is not moving so it can then work form there and teach itself not to it again.
Which CAD software is used to design the 3D model ???
I think you should isolate the camera on some of those little rubber ball that drone's use on there camera rigs. The vibration of those tracks and the frame rate might just meet at the wrong time and the computer can't tell what it is for split second..... easy to test that theory. ...while I got your attention. Put that on the 2 wheel Sonic and teach it the cones and turn it loose.....
Have you thought about using TensorFlow? I've had pretty good experiences with it and allows you to use pre-trained models. I don't have a powerful PC, but training doesn't take more than an hour usually.
Well this is TensorRT
@@jamesbruton oh, I didn't realize.
Can we please send you a Pixetto to see what you can build?? We're so curious!
(11:44) I am surprised the curtains did not mess up the AI.
That's why I had white backgrounds - so it has to see the black shape on white.
I cant get my raspi v2 camera to work on my nano. Anyone have any tips?
It usually doesn't have to take 100s of hours. Can it take 100s of hours? Maybe. But you have to be dealing with a ridiculously large dataset and model. The biggest I've trained took about 2 days. Though the most I've trained took about 2-3 hours.
I've trained models that have well over 5000 32x32 rgb images and it took that 2 days.
can you make a video about how to do ros navigation with a webcam? so no lidar, or maybe how to do navigation with an Xbox Kinect? When are you going to integrate navigation into your sonic hedgehog self balancing robot? I am excited to see that run. I am making a self balancing robot and trying to add SLAM.
Maybe one day but I don't know how to do that myself just yet
Hello: I am a newbie to learning A.I and would like to know what single board computer that I should purchase.
I would like to purchase the Jetson nano 2 gb but it looks like it won;t be available in Canada until next year.
Should I purchase the Jetson Nano development kit with 4gb ram or Raspberry pi 4 with 8gb ram?
Jetson anything is quicker as Neural Nets / Deep Learning
You should make a gen 2 synth robot from fallout 4
i wonder if 3rd printing will drastically increase our produced plastic waste when it becomes more available for everybody
Great project I want to make contact to discuss another project please
I always think of military applications and adding guns.
Today's me birthday :)
Another cracking video thanks. I have the pedant gene, so sorry about this but ... SIMPLISTIC is not just a fancier version of SIMPLE. It means "over-simplified in a way which loses the original meaning". I took the first few with a fixed grimace, but ... you're a Brit! And I can't fight genetics.
Will a Nano 2gb model accomplished this?
You completely missed the differences in approach angles, and in perspective. The initial approach of a square or circle at 45 degrees would see rectangles and ovals, and miss the target.
THEY DON'T ALWAYS LOOK THE SAME -- you keep saying that.
Also, why didn't the shower curtain with the circles get detected at all?
No white background I guess.
Ahhhh
Could you have done the same thing using raspberry pi? (Genuine technical curiosity)
I am a 14 year old boy.
Can u please tell which software you used to make that 3D view of project at 0:47
Fusion 360
@@jamesbruton Thank you for replying genuinely!
@@jamesbruton And sorry for disturbing u again.
I cannot sign up for fusion 360 ,SO can u pls tell any other app for this.
Please reply!
😁
This is a long way off from a virtual girlfriend. But it's a start.. it's a start.
It's probably already more sympathetic, understandable and affectionate than a girlfriend.
with the prices of the realistic robo girls nowadays...pfff.. a robogirl-diy-kit would get a lot of backers!
This comment is trash
Tradução para o Brasil
Did he just predict squid games? XD
👍❤️🌹
No entendi nada pero nice un carro :)
I thought AI was always learning. Or is that a different type of AI? This AI is train then go. That robot should learn while going so it can have the mind of its own 😂
AI - is any program that appears to have intelligence while machine learning actually learns how to get the desired output when provided with an input. In other words machine learning is AI, but AI isn't always machine learning.
@@jesseshakarji9241 ohhh thank you for this!
Actually pretty much all Deep Learning systems train just once. It is a real problem. If you want to add a new object to a trained system then you have to retrain on the complete set of objects again. Systems which can learn on the fly are called 'continual learning systems' and they are still a research area.
@@Paul-rs4gd That's really interesting. This changes my views on AI now! Not in a bad way.
That is NOT a tracked vehicle.... It's a four wheeled thing with a belt.
What's the difference?
@@ChrisD__ How it works mostly.
Can you help i am very disorganized i want to make a robot where it sorts out my tools and puts them back in the tool box but it has to be smart
Messy people always have the better imagination
@@potatotech7378 lmao I have amazing imagination so you can guess how messy i am
The overpowered motors were more fun
Still pretty sad it takes all this to have even less functionality than a small insect
Hmmmm.. this could be trained to 'play' with a cat aka torment it.... nice.
Please remove the task bar (you can press windows key to bring it back), remove libreOffice and download a proper terminal like Terminator.
We all know how coding goes Control c now ur code is copied
I thought you tubers have to disclose when it’s a paid endorsement or at least disclose when they are given things for free by big companies because this is the third you tuber that have praised the jetson nano even tho it’s no more powerful then a pi and is basically a raspberry pi compute board with a normal interface
I bought this one, the Xavier NX is coming up in the next video with disclosure. Jetson runs GPU based stuff quicker than Rasp Pi
James Bruton oh that was a fast reply, I didn’t mean to sound like I was calling you out or anything I just found it weird, James from the hack smith said the same (a bit more defensive but still) but my stance was that you can over clock the pi really easily and with all the heatsinks and hats that you can get for a pi I would argue if you were starting into ai/computer learning development you probably already have a raspberry pi (maybe not a 4 with max ram but still) and with the pi4 having more ram then the nano, they’re about as powerful as one another
And if you were just getting into it but then decided you wanted it in a different project a pi would probably be the better way to go, in my opinion
Thanks for replying tho a lot of people would just ignore comments and answer questions on Instagram or Twitter
Everything is sponsored in this video... It makes me feel gross watching it. The filament is sponsored. Nvidia gave the processing power and there is a general sponser. I think the thing I have a problem with the most, is that you are not making these videos because of your passion anymore. Your only making them for money. Blegh. Have to unsub I am afraid. I really liked your content