Omg I just got it as recommended videos and it was soooo gooood, I was really surprised by seing you only have 600 subscribers! I hope to see some more videos :D. Also thanks for showing the real process of running a project and the difficulties in the way :)
the training dataset is key! for it to work you should do a few things. first try training the s-p6 network. second you need variety. try to put woldo on many different backgrounds. random noise, a beach, a book, lots of things. that will tell him at what to look. else he learns that it has to be on a white background. also if all the training images are perfect digital images he will struggle when you present real photos. include all scenarios in your yolo training and it will work. ;)
Thank you for the insight! What is the s-p6 network in this scenario? I did try to get a couple different scenarios but I suppose I need to get more styles of books and maybe cut a version of waldo and put him in these different situations haha. Thanks for the feedback though!!!!
@@EamonMagd when you specify the model to use, you can use the s-p6. its the S network (small) with an adittional pyramid layer. (6 instead of 5). here is a sample call for the training, that needs to be adjusted for your values: !yolo detect train model=yolov8s-p6.yaml batch=55 epochs=1000 data='/media/cnnserver/Data/mydataset/dataV8.yaml' device=0 cache=true patience=0
@@EamonMagd also note that the Size of wally is imporant! if you train it only with a huge wally it will onyl find huge wallies.. tehs ize should match the searched object or contain multiple different sizes. for the actual search, i would split the images in smller images (640/640) where each image overlaps a little bit to avoid cutting wally in half, and would process them all. should be quite easy and quick in python
Alright, here at 4k views, this is amazing
oh hey, i got pinned!
Your BEST video so far - keep up the good work Eamon 🙌
Thank you for the support as always !
@@EamonMagd You’re more than welcome, Eamon 🙌 Great to see your success 🎉
HOW IN THE WORLD THIS HAVE 110 VIEWS, THIS IS CRAZY GOOD
yeah I know it's missing a k right after the numbers of views ^^
Thanks both I am working on more stuff so hopefully it builds up!
Also annoying commentary 😢
This is super high quality, nice videos man!!
babe wake up new eamon magd video
Yeahhh budyyyy
Wait, the stick figure, the voice…….GRADE A UNDER A?
very interesting! subscribed
Thank you for enjoying! More to come
That was fun to watch. Surprised that it's a small channel of 824 but i just subbed you and now you're at 825.
Waiting for more such videos in future.
Thank you for enjoying the videos! Many more to come man :)
Awesome video!
Glad you enjoyed
Omg I just got it as recommended videos and it was soooo gooood, I was really surprised by seing you only have 600 subscribers! I hope to see some more videos :D. Also thanks for showing the real process of running a project and the difficulties in the way :)
this deserves more views 🤣🤣🤣🤣🤣
this is what the world needs
Where did it waldo wrong 😅
vid was fire
Thanks for enjoying :)
It would be way cooler if the robot had a webcam on it and actually had to move it around to look for Wally.
Ohhhhh that’s a cool idea for the future!
Coriander is the seed. You need to wait a bit before it becomes cilantro.
in canada its the books called waldo i never seen wally befor lol
In Canada, it's also "Où est Charlie?"
liked, commented , very cool video
Big thanks
the training dataset is key! for it to work you should do a few things. first try training the s-p6 network. second you need variety. try to put woldo on many different backgrounds. random noise, a beach, a book, lots of things. that will tell him at what to look. else he learns that it has to be on a white background. also if all the training images are perfect digital images he will struggle when you present real photos. include all scenarios in your yolo training and it will work. ;)
Thank you for the insight! What is the s-p6 network in this scenario? I did try to get a couple different scenarios but I suppose I need to get more styles of books and maybe cut a version of waldo and put him in these different situations haha. Thanks for the feedback though!!!!
@@EamonMagd when you specify the model to use, you can use the s-p6. its the S network (small) with an adittional pyramid layer. (6 instead of 5). here is a sample call for the training, that needs to be adjusted for your values: !yolo detect train model=yolov8s-p6.yaml batch=55 epochs=1000 data='/media/cnnserver/Data/mydataset/dataV8.yaml' device=0 cache=true patience=0
@@EamonMagd also note that the Size of wally is imporant! if you train it only with a huge wally it will onyl find huge wallies.. tehs ize should match the searched object or contain multiple different sizes. for the actual search, i would split the images in smller images (640/640) where each image overlaps a little bit to avoid cutting wally in half, and would process them all. should be quite easy and quick in python
Next time use the project savador extension in Fusion 360 it makes sketches out of images. It saved me a lot of time
I will keep a note for next time 🫡
nice vid man
Cheers man!
What the hell is coriander?
What cilantro (uk version) tries to be mate
I love you
Maybe calm down on the random yelling
bro try To fix the audio
Overedited in my opinion.