What I learned from CodeParade: The most important feature of a face is shirt color What I learned from CaryKH: The most important feature of a face is lighting
Looks scary, but I think it's actually a midway-point between having sunglasses and not having them. Squint your eyes to blur your vision, and the once-terrifying face now looks too cool for school
@@JazevoAudiosurf Actually, became is the past tense of become, so your grammar correction is incorrect. BecauseBGGAMING Deluxe said "has officially", being past tense, the word became grammatically fits. Why don't you actually look things up before you make grammar corrections, so you don't look stupid. If BGGAMING used the word "become", then it would be present tense, and the phrase "has officially" makes the sentence past tense.
I think that the 6th slider is not related to temperature. It is related to being outside or not. Notice it gets redder when the eyebrows go down. There's so much sunlight that the stars slightly squint their eyes. Non-natural light like a light bulb (mostly white) doesn't blind you that much, doesn't hurt your eyes, so you can keep your eyebrows raised with no problem.
Too bad you couldn't ask some random crowd of people to submit clearly labeled pictures of their own faces from a specific angle, with variations where they try their best to match an expression from a standard chart. That could be useful.
Honestly, I think I've learned more about data science, machine learning, and neural networks from watching a few of carykh's videos than I learned from my college Data Science and Neural Networks classes combined. Your "layman's terms" or "basically" explanations are more informative than the wall of equations and mathematical definitions that my professors have tried to use to explain this kind of stuff.
I think the two different traits, PCA 3 and 4 weren't horizontal shading as much as they were brightness for the left and right sides of the screen. While PCA 1 described the overall shading, and PCA 2 described the relationship of the shade of the skin to the shade of the background, I believe PCA 3 and 4 described the relationship between the shade of the skin to the shade of the background to the shade of left or right side.
I work with very low resolution images of 80*60 thermal images. And blurring them with your eyes or looking at them in the distance makes the pixels disappeared and your brain interpolates a good image.
@Tristan Sachsenweger everything you do sensoriell is logarithmic to your brain, due to the nature of neurons - so guess would be it is as well - but once would want to do do experiments.
@Tristan Sachsenweger The brain is used to fill up missing information with experience, sometimes even a few pixels are enough to recognize an object. But if you look on a high res picture with open eyes, the brain expects to see such a high res image and stops interpolating. So if this image is blurry the brain recognizes the blurryness now. If you go to a distance and watch it again, the unclear picture looks like a clear picture in distance, so the brain starts to fill up the gaps again.
Carykh, as long as i've been watching your channel, i have to say you tought me alot, you have made me chukle, and very inspired to do all these things myself, they turned out very fun, it's amazing how much it's been done for all this time. I was hoping to write to you in a long time. Best regards.
I have some KH's that you can use. kept hiding knows humble kisses hammers kangaroo holder kindled hero kind hugs just a few for you to use if you want
You should add a thing so you can put an image in and it changes the sliders by itself to try and match the image. Then we can store peoples faces in only like 50 bytes, as long as we have the program already.
If you want to replicate the Image perfectly, you will have to store as many information as before (plus the program) as a pca is only a rotation if you don't do dimension reduction. You'll have to accept (quite!) some loss of quality if you want to scale it down to 50 dimensions (from the eigenvalues I can see in the video).
0:10 How did this HAPPEN?🎵 A long time ago- actually never, and also now, nothing is nowhere. When? Never. Makes sense, right? Like I said, it didn't happen. Nothing was never anywhere. That's why it's been everywhere. It's been so everywhere, you don't need a where. You don't even need a when. That's how every it gets.
Finally some amazing content I can watch... Honestly been waiting for so long Also I wish some of the stuff like 'evolution sim' were compatible on mic.
I find it quite interesting how the slider will change one main thing, and then one small random thing that you would never link to the main thing but once you start looking further you realise for some reason they are linked!
I think that the ‘math’ may have accidentally stumbled upon some obscure genetic trait, or common combination of traits in humans, in a few cases. When one of the sliders effectively controlling skin color was changed, I noticed that the nose was turned upward as it got darker. Is there some way to further select for a neural network with more sharply isolated traits expressed in the sliders? Maybe something unconventional?
Wait what are the sliders literally just a basis for an eigenspace? Is this what people meant when they said that linear algebra is useful for "machine learning"?
Yes, though there's also some statistics involved in PCA as well. Neural networks require a butt-ton of matrix multiplication and (depending on the training method) some multivariate calculus.
Had linear algebra in my first year and didn't think much of it tbh (was too theoretical for me to image any sort of practical applications) then comes my 3rd year. Data acquisition? Boom! Linear algebra! Image processing? Boom! Linear algebra! Systems modeling and simulation? Boom! Linear algebra!
Dark skin, dark room - *YEA, BOIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII*
the ai was like me at homework, "so we got some pretty got things like temperatur, MakeUp, Smile, Vertical rotation and horizontal shading. But we still 6 more. LETS ADD MANY MORE HORIZONTAL SHADING!!!!!!!"
Great and fun video! It's intuitive that PCA components capture variations in pixels when PCA is done directly on the raw pixel space. But generally it's not obvious that the way the abstract (encoder) version of PCA ranks the levers has a light-related interpretation. You'll notice when doing PCA on the pixel space that you get linear combinations that no longer look like faces, but instead more like blurry superpositions of multiple faces. What the encoder representation buys you, is the ability to move along one encoder axis and stay within a set of what we'd consider faces. Then it's easier for PCA to discover (as you demonstrate) levers of structural facial variation. And that's because the encoder learns to an extent to disentangle what goes on in the lighting from what goes on in the face. Yet, light configurations still seems to dominate the abstract representations. As you move down the hierarchy, facial configurations are more and more salient.
It took so many sliders for it to make sense of 3D! I still cant believe the computer managed to make sense of 3D only provided evidence of its existence, and having no means to understand it without this data, it thought up inky depth to make sense of the missing piece!
What I learned from CodeParade:
The most important feature of a face is shirt color
What I learned from CaryKH:
The most important feature of a face is lighting
Yeah you're right, still that's because CodeParade's dataset actually included shirts in the picture. Cary seems to have cut those out.
I also believe the similarities of yearbook photos also impacted the outputs or the “most important features”
Would be a bit more accurate to say: the most important feature in a *picture of a face* is lighting
horizontal lighting*
codeparade: fursona generator
carykh: objectsona generator (although technically he already did this with bfdi screenshot generation)
11:29 OH GOD WHAT IS THAT
Looks scary, but I think it's actually a midway-point between having sunglasses and not having them. Squint your eyes to blur your vision, and the once-terrifying face now looks too cool for school
That's Joseline Hernandez. It says so at the bottom.
What will happen if you combine Baldi and Principal of the Thing?
Looks like someone who’s either really goth or really tired.
*DEMONIC CHUNGUS 🅱️eter griffin.jpeg*
Awesome! Its so cool to see how different datasets change the different PCA components!
CodeParade you better get pinned
Noice. codeparade is here.
CaryKH stole your idea.
*CAN WE COPY STRIKE CARY!?!*
You mean... “WOW!”
*horizontal shading*
1:37 hands down such a satisfying animation lol
Yea
Yea
Yea
aeY
Not rly
*_Hollywood wants to know your location_*
Your profile is a big dumb meme
V Click this go give m a like
@@M3KAI5ER44 I agree
I had to sub...
Subdemigod summon with so much shit posting.
Congratulations. Carykh has officially became a god.
become*
@@JazevoAudiosurf Actually, became is the past tense of become, so your grammar correction is incorrect. BecauseBGGAMING Deluxe said "has officially", being past tense, the word became grammatically fits. Why don't you actually look things up before you make grammar corrections, so you don't look stupid. If BGGAMING used the word "become", then it would be present tense, and the phrase "has officially" makes the sentence past tense.
@@MyChannel-wp2yd actuSHAULY, the past participle (become) is the correct use here.
ua-cam.com/video/nBcZGjxnpDY/v-deo.html But can he do _this???_
@The Neural Knight i dont have a long enough neckbeard for that.
10:57 “Hey! You look like Michael Huang.”
...
HMMMMMM
doesn’t even look like him lmao
smh
@@bottlecapped it does in another universe
"Eye size"
...
HMMMMMMMMMMM
Spoiler: horizontal shading
Spoiler: other horizontal shading
Spoiler: Smile
I think that the 6th slider is not related to temperature. It is related to being outside or not. Notice it gets redder when the eyebrows go down. There's so much sunlight that the stars slightly squint their eyes. Non-natural light like a light bulb (mostly white) doesn't blind you that much, doesn't hurt your eyes, so you can keep your eyebrows raised with no problem.
“If my brother and I had a baby together”
ALABAMA 100
SWEEEEET HOOOOME ALABAMA
Oh shit so people from Alabama don’t just bang their sisters and cousins? You know what it don’t surprise me
...how mad would your parents be?
@COLD WATER FTW To all the haters who said gays can't have kids. Alabama 💯%
gay 100
alabama 100
12:23 when you stir the nesquik
Joke: Using AI to make evolution
*Woke: Using AI to make p e o p l e*
To all the haters who said gays can't have kids... Pfff... Evolution-checkmate.
@@david2618 That's biologicaly impossible. Proof or STFU.
@@david2618 I saw someone who said he saw a UFO in another man's ass.
_Believe me, i don't have proof._ 🤣
@@rachelslur8729 Oh it was a *joke* and you got *salty*.
Now: Using AI to replace everything
1:26 Them soundeffects tho
The sounds are *burp* an AA
OOF
uur uur UH UH UH UH UH UH UH UH uur UH
**mom walks in*
“We can create a dark-skinned person in a dark room”
Me: I can explain...
Top 10 Snapshots Moments Before *Disaster*
LOL
(Sweats furiously)
oh god
10:57 the algorithm reocgnizes this face as Michael!! I wonder what slider values would recreate Carry.
Drinking challenge: Take a shot every time Cary says horizontal shading
or smile thingy
Drinking challenge: Take a shot each time Ken makes a comment with likes. I am not responsible for any death.
Take a shot for every sponsored video.
Takes a shot 1000 times
I’m nut drank yoir deunk
Col 912 they need to catch dot unicorn on the roof BBQ s NM bgdyiAw
Too bad you couldn't ask some random crowd of people to submit clearly labeled pictures of their own faces from a specific angle, with variations where they try their best to match an expression from a standard chart. That could be useful.
lmao
*Code Parade wants to know your location*
guess who just left a comment
KEn Is MelMoN
Hi
Honestly, I think I've learned more about data science, machine learning, and neural networks from watching a few of carykh's videos than I learned from my college Data Science and Neural Networks classes combined.
Your "layman's terms" or "basically" explanations are more informative than the wall of equations and mathematical definitions that my professors have tried to use to explain this kind of stuff.
10:01 that was literally the funniest thing in the whole video
Oh god you are right
I’m getting out of here! I don’t wanna die! I’m too young to be snapped out of existence!
Half of the pygame is gone now
Thank motion smoothing for dat
Maybe look about 38 seconds ahead and tell me if you change your mind. ;)
2 seconds in and there’s incest!? C’mon.
SupernalZeus238 y e s h o m o
hhhaahahahaaaahahahahaa funi mmeeme
SWEET HOME ALABAMA!
Sweet Home Alabama
@@tapashalister2250 shit
11:16
Cary: "Chew your food Timmy! Or else you'll swallow too much like I shoulda done!"
Me: *Collapses*
*When he changes skin tone and the nose gets bigger*
Oh yeah yeah
go to 11:29
Facts don't care about your feelings
Yeah I looked more masculine with a darker skin tone.
Edit: It's supposed to read "it". 🙈
@@GhostSamaritan No you don't. Green is the new gangster. Ree
13:00
You blurred it! Carykh is a *legend!*
“If my brother and I had a baby together”
*pauses video*
wh-
@@nhehguy nothing can be more accurate than that, thank you for your service.
Sweet home Alabama
I think the two different traits, PCA 3 and 4 weren't horizontal shading as much as they were brightness for the left and right sides of the screen. While PCA 1 described the overall shading, and PCA 2 described the relationship of the shade of the skin to the shade of the background, I believe PCA 3 and 4 described the relationship between the shade of the skin to the shade of the background to the shade of left or right side.
I work with very low resolution images of 80*60 thermal images. And blurring them with your eyes or looking at them in the distance makes the pixels disappeared and your brain interpolates a good image.
@Tristan Sachsenweger everything you do sensoriell is logarithmic to your brain, due to the nature of neurons - so guess would be it is as well - but once would want to do do experiments.
@Tristan Sachsenweger The brain is used to fill up missing information with experience, sometimes even a few pixels are enough to recognize an object. But if you look on a high res picture with open eyes, the brain expects to see such a high res image and stops interpolating. So if this image is blurry the brain recognizes the blurryness now. If you go to a distance and watch it again, the unclear picture looks like a clear picture in distance, so the brain starts to fill up the gaps again.
Carykh, as long as i've been watching your channel, i have to say you tought me alot, you have made me chukle, and very inspired to do all these things myself, they turned out very fun, it's amazing how much it's been done for all this time. I was hoping to write to you in a long time. Best regards.
bUt It _Is_ fEbrUAry 2019!1!!!!
h u s h
it means that part 2 is coming at max, at the end of February
“How much horizontal shading do you want?”
This program: Yes
10:06 you just made Steve Harvey LMAO
1:36
... I could watch this part In a loop forever... So trippy and so satisfying
I have some KH's that you can use.
kept hiding
knows humble
kisses hammers
kangaroo holder
kindled hero
kind hugs
just a few for you to use if you want
Also klutz hitter
Kill Hitler
Knowledge hope
Actually makes sense
Cary Key House
Cary keep hididng
Cary king horse
Cary killed holmes
More kh
Kicked hitler
Kills humans
Kind human
Koopa Hider
Key Holder
Kite holder
Kale heater
That’s all I know
The good old days where generative AI was used only for fun and not to replace artists
i miss those days
It’s not BFB 13 but it’s still art
If he used it on all 63 BFB characters it would be the face of team blcsafdtop
0:00 “OH MY GO-“(beep)
I WANT TO SEE THE PICTURE A
I SAW THE ORIGINAL HE EDITED IT OUT sadly no image tho
So like that's something...
*i have several questions*
Ah, thanks, I wouldn't have seen it if you hadn't mentioned it!
I’m now curious how Michael and Cary would have a baby...
lol me too
3:20 is the "yeah boi" guy
The world’s greatest character creator
1:28 Me When I Chew 5 Gum
*burp* and AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
Really cracked me up
New 5 Gum. Stimulate your senses.
What was that beginning.
What was that beginning.
I'm scarred for life.
What if Kim Jong Un and Joseph Stalin had a baby?
Skillshare and Brilliant
It will be OUR baby
ThIS ViDEO iS SponS0r3d b6 sKiiLSh4r3 aÑð ьгїlLįæñT uSE c0d3 wWw,SkiLLsH43XþrìlLÎaÆnjț.COm.0rG\HāÏ
Oh wait... Wrong video... Pardon me, sir.
@ThatOnePlayer 2019 wha... What?
i was scared of the beginning but luckily you changed it
10:57
Wow, thats how close brother ship can get.
Woah, you mentioned the two most recent presidents without making a joke about politics, good job. 👍
political Humor XDDDDDDDDDDDD LMAO JAHAHAHAHAHAHAHAHAHAHAHAHAHAHHAHAHAHA
Is D.R.T. (initials) still our president?!
1:27 is the best part of any carykh video
You should add a thing so you can put an image in and it changes the sliders by itself to try and match the image. Then we can store peoples faces in only like 50 bytes, as long as we have the program already.
If you want to replicate the Image perfectly, you will have to store as many information as before (plus the program) as a pca is only a rotation if you don't do dimension reduction. You'll have to accept (quite!) some loss of quality if you want to scale it down to 50 dimensions (from the eigenvalues I can see in the video).
@@LegoEddy I understand it won't look great, but it would just be fun to mess around with.
Can we get a 10 hour loop of that demented Jeopardy song that gets worse as it goes?
Cary and Michael having a baby, now the shippers are gonna go insane.
w what shippers. who the fuck would ship brothers
0:10 How did this HAPPEN?🎵
A long time ago- actually never, and also now, nothing is nowhere. When? Never. Makes sense, right? Like I said, it didn't happen. Nothing was never anywhere. That's why it's been everywhere. It's been so everywhere, you don't need a where. You don't even need a when. That's how every it gets.
This is boring
@@tessaallan8478 forget this
OMGH
@@rubiks_20I wanna be something, go somewhere, do something
@@tzabarpho.i want things to change, i want to invent time and space
4:09 i swear why did that make me laugh
13:18
Cary: So at some point, you'll have to turn to this video's sponsor, brilliant.org!
Me: Wait, WHAT???
Did just cary used the anime wow sound effect?!
0:49 If you're doing this like CodeParade, than are you also making a video on Computer Generated Fursonas? Oh no...
who makes the best fursona?
Yes please
@@kellynyanbinary I like your atitude!
I think he should make a trap generator. I mean, he uses that anime sound effect
@@PixelBytesPixelArtist That's a yikes from me dawg! Not saying it's not a good idea though...
9:04 got me laughing 😂
Got me cracked up
I was holding in my laugh so hard
Finally some amazing content I can watch...
Honestly been waiting for so long
Also I wish some of the stuff like 'evolution sim' were compatible on mic.
Very interesting! Can't wait to see more of this!
Carykh You Think We Didn't See The True Jesus Of This World Hatsune Miku! 11:57
*Miku
Stupid Ayto Corect
11:26 looks like Tom Scott😂
I'm dead 😂
12:44 scary I recognized the sound effect from hunterxhunter
10:57 Whao didn't expect Michael to be there
1:09 The arrival of the fabled: John Fame
"woah Thanos what are you doing here you weren't invited now I don't feel so good" captions
"It would be fun to ad a joystick control. But that's too much work!!!"
*don't worry dude I know how that feels...*
I’m glad you got a sponsor Cary.
2:32 It's trippy how it looks like the nose gets bigger when the image gets darker though
I find it quite interesting how the slider will change one main thing, and then one small random thing that you would never link to the main thing but once you start looking further you realise for some reason they are linked!
This would be a great tool for police sketches and describing people.
Also, glad you uploaded!
I think people are already doing this
Describe people,
You know, he was that guy who has 1.314 4.24 8.24 48.29 35.4 91.3948
It's really cool of you to promote CodeParade
4:30 - The background's dark, almost black.
4:40 - We have a light colored background.
I also noticed that 4:30 smiles more than 4:40.
I think that the ‘math’ may have accidentally stumbled upon some obscure genetic trait, or common combination of traits in humans, in a few cases. When one of the sliders effectively controlling skin color was changed, I noticed that the nose was turned upward as it got darker.
Is there some way to further select for a neural network with more sharply isolated traits expressed in the sliders? Maybe something unconventional?
The "obscure genetic trait" known as ancestry lol
aka race is mathematical true? i saw it also for the temperature where low temp moved the eyebrows up
If you have labelled data you can use Conditional GANs
"math doesn't know what it's like to see color" damn guess im colorblind
Im sorry about you knees, praying over here, hope you recover.
Wait what are the sliders literally just a basis for an eigenspace? Is this what people meant when they said that linear algebra is useful for "machine learning"?
Yes; and partly, but linear algebra is also required in many problems as a handy tool.
Yes, though there's also some statistics involved in PCA as well.
Neural networks require a butt-ton of matrix multiplication and (depending on the training method) some multivariate calculus.
Yes, machine learning is literally A LOT of linear algebra
Had linear algebra in my first year and didn't think much of it tbh (was too theoretical for me to image any sort of practical applications) then comes my 3rd year. Data acquisition? Boom! Linear algebra! Image processing? Boom! Linear algebra! Systems modeling and simulation? Boom! Linear algebra!
I can’t wait for part 2!
11:29 Is it just me or do they look like they're going to break into your house and sacrifice you to satan
8:22 "you look like Onision"
kill me... NOW.
5:22 GD BROTHERS WHERE U AT
0:00 He said Boom Mic
Cary Huang: Oh my boop
What I noticed that he said was Boom Mic
Boom Mic from S5: You called me?
Dark skin, dark room - *YEA, BOIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII*
lol
That expanding branches animation is just trippy
8:35
IS THAT MARINA AND THE DIAMONDS
AAAAAAAAAA
10:57 WHAT? MICHAEL HUANG?!?!
10:44 hit or miss
i guess they never miss, huh?
I love that the makeup slider made the bottom text say 'I look like onision'
11:52 had me dead lol
This one is actually very impressive. Good job.
1:27 My sense of humor has evolved so much that I can’t stop laughing at this part
This is getting really professional
the ai was like me at homework, "so we got some pretty got things like temperatur, MakeUp, Smile, Vertical rotation and horizontal shading. But we still 6 more. LETS ADD MANY MORE HORIZONTAL SHADING!!!!!!!"
Great and fun video! It's intuitive that PCA components capture variations in pixels when PCA is done directly on the raw pixel space. But generally it's not obvious that the way the abstract (encoder) version of PCA ranks the levers has a light-related interpretation. You'll notice when doing PCA on the pixel space that you get linear combinations that no longer look like faces, but instead more like blurry superpositions of multiple faces. What the encoder representation buys you, is the ability to move along one encoder axis and stay within a set of what we'd consider faces. Then it's easier for PCA to discover (as you demonstrate) levers of structural facial variation. And that's because the encoder learns to an extent to disentangle what goes on in the lighting from what goes on in the face. Yet, light configurations still seems to dominate the abstract representations. As you move down the hierarchy, facial configurations are more and more salient.
0:25 Wow. Really gonna do my boy John Smith dirty like that, huh?
It took so many sliders for it to make sense of 3D! I still cant believe the computer managed to make sense of 3D only provided evidence of its existence, and having no means to understand it without this data, it thought up inky depth to make sense of the missing piece!
it s almost as if math had consciousness
just came back from a video explaining all the types of NNs and now I actually understand what Mr. KneesHurt is talking about
10:53 looks like baldi
Congratulations! You've made a nightmare generator
2:52 "Blueface baby, ye aight"
I dont understand half of the stuff here, but i still love your videos. It really says alot about the quality of them ;)
That's how AI works.
Sam Montanaro is that a “that’s how mafia works” reference?
im a tree yes
11:56 - Ah, I see you're a man of culture as well!