Thanks Nicholas for this video, I am currently working on my master thesis which is something like this. It's nice to see you doing it too, it gives me some new interesting ideas and knowledge! Thanks again and keep it up!
While watching I was wondering, did you pass the raw joints' point positions to the ML algorithm or did you do any sort of feature engineering and pass those features to the ML algorithm?
Great question, I didn't end up doing any FE but that was my next step if I was going to take this further. I definitely think providing features with relative distances e.g. x1-x7 would help a ton when analysing specific movements. I'm going to share the full code this weekend so you can take a look!
Nick, I cannot thank you enough for sharing your priceless knowledge! On video descriptions of your previous videos, you said that people who need some help for their course, startup, or business can get some help out of your video and you opened up your codes. So you allow the viewers to use your codes/materials? I also would like to use some of your materials for my school project. Of course I will be referring to your video/sources and clearly mention it on the reference page, but is there any proper way you want us to cite your materials?
Definitely go for it, just make sure you reference the code used so you don't get in trouble with school or follow their guidelines. I'm cool with you leveraging the code! I'm going to be sharing the source for this, this weekend!
@@NicholasRenotte Thank you so much for your reply. Your works are wonderful, but your bright personality and fantastic teaching skills made them way easier for a beginner like me to understand. I appreciate it.
So sick. Nice work Nicholas! This is such a great idea. Tracking form (& correct form) can be used in so many applications. Can you do this again for basketball shots? Golf swings?
Is there anyway you could do an in depth tutorial on this? Or at least how you were able to get MediaPipe working on your raspberry pi? I've been having trouble getting MediaPipe to work as "pip install mediapipe" won't work. *Please help if you see this.* Also thanks so much, I've been learning so much from your videos!!
Hi Nicholas... Amazing job!! Thanks for sharing... By the way, why don't you try out this implementation using the Nvidia Jetson Nano? It's a little bit expensive but has more performance and it's becoming more popular between the makers that use edge AI. See ya!
I wish I’d bought one for this, would’ve been the perfect application! The UI and models run at light speed on the dl rig, will probs get a nano to test this week!
@@NicholasRenotte great idea! i was thinking same direction to try Nvidia Jetson Nano or Google Coral Dev Board. Alternatively trying to enhance RaspberryPi with Intel Movidius stick. Please share a vid if you try one of those. Anyway, kudos for a great work man! cheers
Hey, I really want to build a similar app as a personal project. Can you provide some insight into the parts you didnt cover in the video, like how did you clean up your data to make the predictions acurate ?
could u lease drop the code and steps for this amazing project it would be so much helpful i am a university student and i am working on a similar project
Hi Nicholas ,ur videos helped me a lot ...ur series of object detection videos are one of reason to crack 5 offers in my hand now.If possible can u please made series of videos on MLops and AIops and deployment strategies because I am not able to get perfect path to learn this skill set.
Awesome video. Easy to understand for beginners like myself! I am currently doing a university project about this, so I was woundering if it's possible to get access to the code? :) Once again, thanks for the video.
Hey Nicholas, thank you so much for making these videos. I work at a company which does fit AI and these help alot. Could you do a video where you're doing signal processing for finding out the errors done during an exercise? That's the solution to most of the dynamic exercises and I am struggling with it a little bit. It'll help out alot. Cheers man. Great work! 👍🏻
@@NicholasRenotte that'll be great man. Would help out alot actually! Or if you could share the python scripts you used to build this, that would be great too!
@@NicholasRenotte yeah that would be great :) but it seems as though your area of interest have changed a bit. still would make great educational content.
Nicho,I am a big fan of you!And i got some inspirations from your road of becoming a data scientist that i also wanna record videos when i finish learing a project by myself.Could you give me some guidance on basic preparations and knowledge before recording a video?
After watching all of the videos in your channel, I am able to make my own AI trainer that has a lot of functions like being able to speak, send performance via email, help us to see whether your body is correct when doing exercise etc. many thanks to u! One of my biggest inspiration to study CS at University!
Hey Andrew, could you please share the code to your AI trainer. I am working on something similar and it'll help out alot to see what you've done and correct my mistakes.
Excellent.. Thanks Nicholas..can you please extend this to detect deadlifts for multiple person in a video..I think you need to have one for action detection in a video having more than one person.
hi nicholas, i followed your videos and was able to detect handwriting and display their labels above the box. however, im stuck on trying to order the boxes from left to right and then top to bottom. any idea or suggestions on how to proceed? Any help would be much appreciated. Thanks!
Use the top right most coordinate and order from there. So look for the smallest x and y value then iterate based on growing x then by growing y (depending which way your y axis is ordered)
hi nick. I'm curious how you optimize the model? to get the result better, did you delete the key point that are not necessary or you do something else?
Great tutorial! I have question. Could you share how labels are consist of? There are columns named x1, y1, z1, v1. I think x,y,z is the coordinate of every joint but i have no idea about 'v'. Thanks!
it's really very helpful video for my project. if possible could you make video how to train the model, annotate and clear the data and load to the model.
loved this one :) came back to your channel cus i was working on something i knew you mentioned something about it in some earlier vid found byself bingewatching :) well guess i have to digging up my pi again and build this now thanks
This is pretty cool, I must say! Great job will be an understatement here. Btw II was going through the audio classification video of yours and decided to implement the same idea onto the Urbansound8k dataset but there is something about some of the files that are giving an error that only gives me an integer and doesn't explain what could be wrong. Any idea on what could have gone wrong or what can I do to fix this??
Could you please explain how you made the rep counter? It would be very helpful if you could share the code please or make an in-depth tutorial of this. Thanks for the video!
Thanks Nicholas for this video, I am currently working on my thesis which is something like this. It's nice to see you doing it too, it gives me some new interesting ideas and knowledge! Kindly share me the code..
Your videos are amazingly engaging and one can learn a lot from them, they are also very underrated I hope everyone recognizes your vids and have a chance to learn from you. Thanks a lot, keep up the good work!
Hi Nick, I’m new to your channel and I appreciate the videos you bring to us they are interesting informative and teach new innovative ways to learn new projects. I do have a request which is to consider making a 30-45 minute python crash course that covers the basics of python programming/or basics to get us started making projects in python etc
Awesome!!! We need a part two where you work on adding some of those enhancements you mentioned. Maybe supporting multiple exercises etc. Also you mentioned you did a lot of cleaning up with the data that helped improve the model. What exactly did that process look like? Do you have another video on that?
I’m thinking I’ll actually do a big tutorial on it if people want to give it a crack. For data cleaning I tightened up the way I labelled so I would only export coordinates at the absolute top and absolute bottom of the pose.
Is there anyway you could do an in depth tutorial on this? Or at least how you were able to get MediaPipe working on your raspberry pi? I've been having trouble getting MediaPipe to work as "pip install mediapipe" won't work. *Please help if you see this.* Also thanks so much, I've been learning so much from your videos!!
@@NicholasRenotte awesome! Thanks so much!! Also would be amazing if you could lmk how you were able to get Mediapipe to work. I've been struggling lol
Enjoying your videos and your tutorials!! I would like to try the same for other workouts too but it's a lot of work to get the data, pre process it and then train the model for each workout. 😵💫