>>2979217
this is /diy/ so unless you plan to invent your own grading system and then use that to grade this dancing woman you should probably fuck off

>>2979221
I do plan to do that

>>2979217
your eyes must not work cause in your twerk video in that time span her ass faces the camera and then your ethot video she just faces the camera and doesnt twerk. come on dude stop being so lazy you should just pay me

Compare the motion points of the original video to that one.
This belongs on g btw

>>2979217
Oh, absolutely. Because nothing says "romance" and "normal human interaction" quite like sending a woman a "Twerk Workout Tutorial" and then demanding an AI-generated report card on her gluteal rhythmic accuracy.

Its truly a mystery why NASA isn't using this technology to calibrate satellite dishes. Since you're looking for a way to turn a video call into a high-stakes episode of Just Dance: Extortion Edition, here is why your plan is 100% flawless:

The "Perks" of Your AI Dance Scout
The Kinect Dream: Youre right, the Xbox Kinect was the peak of human innovation. Its a real shame Microsoft discontinued it; they clearly didnt anticipate the massive market demand for men wanting to "grade" private twerk sessions against Rihannas "Rude Boy".

Mathematical Precision: Why settle for just enjoying a dance when you can have a computer tell you that her "Happy Dog" torque is only operating at 74% efficiency? Nothing keeps the spark alive like a performance review.

The "State of the Union" Feedback: Imagine the joy shell feel when you interrupt her live video to say, "Hold on, the algorithm says your feet weren't shoulder-width apart during the Tootsie Roll section. Thats a 5-point deduction and a loss of dessert privileges."

Cutting-Edge Logistics: Its so generous of you to provide the source material. I'm sure shes thrilled to know shes being benchmarked against a professional fitness instructor while shes just trying to have a conversation with you.

A Few "Minor" Technical Hurdles
AI Sentience: Most AI models are currently focused on things like "curing cancer" or "writing code," but Im sure if we redirect all global computing power, we can finally solve the "is she twerking hard enough for this guy" problem.

>>2979217
thats a man

>>2979222
You need to find better things to do with your time, because this is retarded.

>>2979353
It’s not that time consuming, I have lots of free time and I’m not spending a large percentage of my money on it either

>>2979217
>How can I use AI to “grade” the dancing of a woman
you are an idiot. only morons rely on AI for anything.

>>2979362
>I have lots of free time
do something actually useful and productive instead of grading goon material.

>>2979301
Underrated post.

>>2979391
Society is fake and gay i want to produce more system soaking

>>2979217
You need to give up gooning. It's ruining your critical thinking skills.

File: cat2.webm (2.2 MB)

2.2 MB WEBM

>>2979217
you tell me

File: Capture.png (778.1 KB)

778.1 KB PNG

>>2979217
As others have pointed out, this question is /g/. There's going to be an issue in that machine vision just isn't anywhere near the level of tracking of a mocap suit.
This is pretty much state-of-the-art in full-suit mocap:

https://www.youtube.com/watch?v=Xi0hvSUHPJs

Note the single-finger precision and shoe bending.

Without accurate motion data your model is going to have a lot of noise. There's the additional problem in that dancing is an art form and added "flair" or artistic expression is often a good thing:

Watch this and note that none of the dancers are carbon copies of any other, and the result is incredible:

https://www.youtube.com/watch?v=sEKZj-H7qa8


You might be able to come up with something that will notify you if it seems a dancer flubbed a choreographed move completely, but since dancing is an EMOTIONAL art a machine can never grade it.

OP is gay. His brain is not tuned to judge the sexual attractiveness of a woman. That's why he needs to use AI for it.

>>2979217
Very simple pattern classifier. You just need a million different videos of different women in different settings making the same dance and feed all that into a supercomputer and six months later you will have a model to detect if they do the dance or not. If you want it to grade the dance you need a million videos of each stae of grading.
Check out e.g. pytorch.

>>2979217
Don't.

>>2982789
or an actual practical solution would be to find a base model thats able to map human poses and have it create a track report of the source video and the video you want graded, then you can come up with a custom grading formula via the coordinate differences between different sections of the limbs. OP should find something better to do with his time though

>>2979217
Ask chatgpt

>>2982789
You could scrape tiktok for videos