This is insane. I think people cannot appreciate enough for putting 7 months into this for a 20min video in the end but man, that was a lot of fun to watch! Awesome stuff!
This is sooo awesome!! A shame it is so underappreciated. You should really present this at Open Sauce 2024, it is the maker exhibit for content creators and fun projects like this. people would LOVE to see it at a convention like that. Seriously, the organizer made the registration fee free for exhibitors
Hey coy this is just awesome! Sad it doesn't have more views yet. Very impressive. Maybe change the title to sth like "this Robot plays Uno for you" for it to get picked up by the algorith. I will send it to all my engineering friends now. Also subscribed, great content. Keep it up. Cheers!
Very cool project you've made! The way you show how semi-complex tools (e.g., ML) can be used for this kind of task is great. This video gives a great understanding on ML can be used with industrial automation. If you wanted to take a different approach to ML for colors. Using the color system CMYK instead of RGB might allow you to create something yourself and save a bit on the computing side (image processing) of the project. You can convert RGB into CMYK with some loss of color accuracy or vibrancy, the conversion mainly hits the very vibrant or saturated colors, as the two color models represent colors differently. Also, you mentioned how only one of the arms could move at a time. It sounds like the program used, gives them instructions in a single-threaded manor, (better for safety in your workspace, where they could hit each other) that could be the reason only one of them would operate at a time.
Thanks, very good points made. The CYMK color scheme would have definitely been better and that is a likely explanation of only being able to control one arm at a time. Thanks!
Very cool project! You put a lot of effort in to this and it paid off however I wanted to point out that it is possible to "over train" a model, take for instance the most popular dataset for characterizing the digits 0-9 , MNIST, it's only about 60,000 images and you can get really good models from it. Also as you already know the training dataset is really important however its just as much about variation as it is repetition, you want to be able to find a specific pattern amongst what is basically background noise, so if you have many copies of the same image with the only variation being brightness or rotation then that repeated background is now part of the pattern, your not only looking for the number your also looking for what is common between all the other images.
I'm not a fan of calling that AI, I am however a fan of you doing this because it is hilarious. I would just like for the humans controlling robots to do illegal turns, like drawing a card when they don't have to, playing +4 when they shouldn't be able to, choosing the wrong number or colour. Just gives it a better feel and forces them to think more about their options, making bad decisions more common.
10:00 I would do this using the so-called 'Euclidean distance'. If we imagine the numerical values of the three color channels as coordinates in 3-dimensional color space, then we can determine which of the 8 basic colors is closest to our color. Simply measure the distance from point to point in color space.
Wish in portugal we can had acssses to thooss robots , i would like to work with them , maybe learn a thing or twoo , but unforntlyt everything is business so no chanche , nice video keep them comming😎😎😎
Portugal is part of the EU, so you can easily buy used robots. In my country it is much more difficult to do this - even used robots are outrageously expensive.
Thanks! One is shorter just showing what it can do. The other has more detail on how it was built. The third shows a complete game and functionality. All together would be too long of a video.
I coded my own in Visual Studio using a Windows Forms Application (C#). There's a lot of if statements in a loop for each turn. That's how the computer is "playing" the game.
He used ai to recognize the cards? And what do you consider AI? Deep blue beat kasparov and that wasn't using machine learning, is it not AI? I think you're confusing ML and AI
This is insane. I think people cannot appreciate enough for putting 7 months into this for a 20min video in the end but man, that was a lot of fun to watch! Awesome stuff!
Lovely automation
This is sooo awesome!! A shame it is so underappreciated. You should really present this at Open Sauce 2024, it is the maker exhibit for content creators and fun projects like this. people would LOVE to see it at a convention like that.
Seriously, the organizer made the registration fee free for exhibitors
Interesting 🤔
This is awesome! I've wanted to get into robotics myself and your videos are an inspiration to me.
Hey coy this is just awesome! Sad it doesn't have more views yet. Very impressive. Maybe change the title to sth like "this Robot plays Uno for you" for it to get picked up by the algorith. I will send it to all my engineering friends now.
Also subscribed, great content. Keep it up. Cheers!
Awesome! Thanks for the support
Very cool project you've made!
The way you show how semi-complex tools (e.g., ML) can be used for this kind of task is great. This video gives a great understanding on ML can be used with industrial automation.
If you wanted to take a different approach to ML for colors. Using the color system CMYK instead of RGB might allow you to create something yourself and save a bit on the computing side (image processing) of the project. You can convert RGB into CMYK with some loss of color accuracy or vibrancy, the conversion mainly hits the very vibrant or saturated colors, as the two color models represent colors differently.
Also, you mentioned how only one of the arms could move at a time. It sounds like the program used, gives them instructions in a single-threaded manor, (better for safety in your workspace, where they could hit each other) that could be the reason only one of them would operate at a time.
Thanks, very good points made. The CYMK color scheme would have definitely been better and that is a likely explanation of only being able to control one arm at a time. Thanks!
This has got to be the coolest thing ive seen in a while. Truly underrated video
Awesome videos. Love your channel, keep up the good work, greetings from Germany.
Very cool project! You put a lot of effort in to this and it paid off however I wanted to point out that it is possible to "over train" a model, take for instance the most popular dataset for characterizing the digits 0-9 , MNIST, it's only about 60,000 images and you can get really good models from it. Also as you already know the training dataset is really important however its just as much about variation as it is repetition, you want to be able to find a specific pattern amongst what is basically background noise, so if you have many copies of the same image with the only variation being brightness or rotation then that repeated background is now part of the pattern, your not only looking for the number your also looking for what is common between all the other images.
I see, that's very interesting and explains a lot. Thank you.
I'm not a fan of calling that AI, I am however a fan of you doing this because it is hilarious.
I would just like for the humans controlling robots to do illegal turns, like drawing a card when they don't have to, playing +4 when they shouldn't be able to, choosing the wrong number or colour. Just gives it a better feel and forces them to think more about their options, making bad decisions more common.
10:00 I would do this using the so-called 'Euclidean distance'. If we imagine the numerical values of the three color channels as coordinates in 3-dimensional color space, then we can determine which of the 8 basic colors is closest to our color. Simply measure the distance from point to point in color space.
Interesting, that could be another way of doing it.
Cool
Nice problem solving skills
Algorithm finally learnt. What amount of work and dedication you put into your projects! Wish you a great career!
Wish in portugal we can had acssses to thooss robots , i would like to work with them , maybe learn a thing or twoo , but unforntlyt everything is business so no chanche , nice video keep them comming😎😎😎
Portugal is part of the EU, so you can easily buy used robots. In my country it is much more difficult to do this - even used robots are outrageously expensive.
What are the specs of the computer you used for the thought process of the computer? Was it demanding or is an old dell tower useable?
To build the model it would take a stronger computer. But to run the model slower, a weaker computer works. But it depends
😍 robots
The amount of dedication put into this 🦾youre really an inspiration man
Nice Video. But why post it three times?
Thanks!
One is shorter just showing what it can do.
The other has more detail on how it was built.
The third shows a complete game and functionality.
All together would be too long of a video.
@@coybeardmore I don't think the algorithm is gonna like it though, might want to just cut out more stuff and make it one video
What simulation software were you using?
I coded my own in Visual Studio using a Windows Forms Application (C#). There's a lot of if statements in a loop for each turn. That's how the computer is "playing" the game.
You didn’t gave it AI, you gave it a lot of memorize schemes. Y’all guys need to let go of that buzzword already. It’s tiresome.
He used ai to recognize the cards? And what do you consider AI? Deep blue beat kasparov and that wasn't using machine learning, is it not AI? I think you're confusing ML and AI