NCAA joins CMU's own investigation of sideline sighting of CS
Even more reason that the Big Ten should just leave this to the NCAA.
November 6th, 2023 at 1:59 PM ^
There is absolutely nothing new in this "report" by ESPN other than a statement from CMU's AD that says CMU "continues its review of the matter in cooperation with the NCAA".
What a bunch of hacks. Just an effort to keep this story on its headlines.
November 6th, 2023 at 2:04 PM ^
Per the article:
Jain said the system produced a similarity score of 0.6 when comparing the two photographs. To validate that score, Jain and Grosz compared Stalions' photo to a database of more than 4,500 photos of white males.
"The reason why it's 0.6 is because there's a disguise," Jain told ESPN. "If I take an identical photo, it would be one. Even changes in the pose, illumination, expression, sunglasses, the match will never be perfect. Based on this analysis, the two images are of the same person with high confidence."
0.6 is not any kind of high confidence. Not saying it wasn't him, but this is just laughable.
November 6th, 2023 at 2:08 PM ^
It's just not clear if 0.6 is a confidence score or not. The article omits any information that would allow the reader to judge its conclusions.
November 6th, 2023 at 2:27 PM ^
I assume that's part of the point - it's probably him but whatever analysis he's using here is coming back with "yeah, he's a white guy with a goatee" and that has a ceiling in terms of correlation/confidence. It's why I found it funny he said this was a "disguise" - it's a pair of sunglasses. Good facial recognition software (which I assume this "state of the art" one is) can work around that pretty well. I'd assume it would be a higher confidence number than 0.6.
Also, I looked the guy up and while the article says "distinguished professor at MSU's EECS" department, his actual title is University Distinguished Professor, which basically means MSU put "distinguished" in his title like a nickname. He's probably a good CS professor but it's weird they tried to sneak that by everyone.
I think it's Stalions but there have got to be better people you can talk to than this one guy at MSU.
November 6th, 2023 at 2:30 PM ^
My iPhone recognizes me with sunglasses on. How hard could it be? Disguise…ha.
November 6th, 2023 at 3:05 PM ^
Yeah, I don't know a ton about facial recognition software but when I was in grad school a decade+ ago I remember a computer vision course that was making decent strides at handling stuff like fake facial hair, glasses, hats, etc. The technology is there, and so if he's waffling here it makes me think his results are a bit wonky but he knows it's Stalions and doesn't want to say his state-of-the-art research is worse than my eyeballs and slight context clues.
November 6th, 2023 at 3:43 PM ^
My eyeballs have it at a 0.5, as in it could be him, and it also couldn't be him.
November 6th, 2023 at 3:37 PM ^
And even with a mask!
November 6th, 2023 at 3:42 PM ^
Your iphone is not just using a 2d camera. It produces a three dimensional facial map using an infrared array.
November 6th, 2023 at 3:15 PM ^
My iPhone recognizes me in sunglasses. And I assume the confidence level (or whatever the score means) is higher than 0.6.
November 6th, 2023 at 3:40 PM ^
Your iphone uses an infrared dot matrix array to generate a three dimensional map of your face. It is not using image processing software on a two dimensional image.
I believe that Samsung and Google use their cameras for facial recognition, and I understand they do a much worse job of it than the iPhone does.
November 6th, 2023 at 2:28 PM ^
You're right that the scale and its operation are not laid out--and here they're just cutting and pasting from a previous article. OTOH, the investigator said that the score was strong enough to indicate great likelihood of a match.
I don't think CMU's protests of uncertainty are going to wear well, though.
November 6th, 2023 at 3:15 PM ^
Nah, I have cracked the code and did an analysis control study.
George Costanza would be 0.5 and Justin Timberlake would have been a 0.7, so it's some where in between there in the degree of certainty.
November 6th, 2023 at 3:38 PM ^
This is laughable. The suggestion that 60% if high confidence is ridiculous. That being said, and this is only opinion, I tend to agree with the theory that CMU was aware that Conner Stallions was really good at deciphering other teams signs and may have hired him. That would explain why he had the pass to be near the coaches. It seems like they are slow walking on their investigation, which is suspect.
November 6th, 2023 at 3:45 PM ^
It's unlikely to be CMU as an institution. That would be very stupid of them. Much more likely one of their coaches, and one with power to issue a sideline pass... in a different name. Recall McElwain's press conference, where he stated he didn't know about it, and the sideline pass was not issued in his name. They were "tracing it back...".
We'll soon find out, I'm sure.
November 6th, 2023 at 3:44 PM ^
Anything .59 or under is "way fucking off". Anything .60 or higher is a match
November 6th, 2023 at 4:20 PM ^
It is definitely him or Edward Norton, definitely. 60% of the time it is right every time! Hang'em high.
November 6th, 2023 at 2:09 PM ^
I have to think MSU’s football team would be slightly more successful worrying about who is on their own sidelines than that of their opponent.
November 6th, 2023 at 2:11 PM ^
“See see the low score just PROVES it’s him because he’s CLEARLY wearing a disguise!!!”
Definitely some conspiratorial reasoning there. Hacks.
November 6th, 2023 at 2:15 PM ^
So all it takes is a hat and sunglasses to throw off the score?
Does that guy with hat and sunglasses possibly match someone else not in your 4500?
Yes, 2 identical photos would be the same. We need your shitty tech to figure that out?
November 6th, 2023 at 2:20 PM ^
So all it takes is a hat and sunglasses to throw off the score?
Worked for superman...
November 6th, 2023 at 3:41 PM ^
I don't think their groundbreaking software will sell very much if a rudimentary disguise can throw it off 40%.
November 6th, 2023 at 4:21 PM ^
Well it did determine it's a white person with facial hair...so...
November 6th, 2023 at 2:16 PM ^
Facial recognition software is extremely problematic
November 6th, 2023 at 2:32 PM ^
ok, so, if i take an IDENTICAL photo and compare it to another IDENTICAL photo, we are confident they are the same photo.
Stellar work.
November 6th, 2023 at 3:43 PM ^
I want to see proof that their software would be able to identify 2 identical photos at a confidence interval of 1.0, when slight alteration of appearance moves it to a coin flip if that was him.
November 6th, 2023 at 2:50 PM ^
"The reason why it's 0.6 is because there's a disguise," Jain told ESPN.
This is blatant "begging the question" and reveals ingrained bias. All the more reason to relegate this pseudoscientific rubbish to the trash bin where it belongs.
November 6th, 2023 at 3:00 PM ^
Is that out of 100. lol
November 6th, 2023 at 3:11 PM ^
I wonder what section of his university workload document Professor Jain filed time spent on this task under
November 6th, 2023 at 3:24 PM ^
"The reason why it's 0.6 is because there's a disguise," Jain told ESPN. "If I take an identical photo, it would be one.
I don't know if this guy assumes we're idiots and is talking down to us, but this is some incredibly dumb reasoning.
November 6th, 2023 at 4:13 PM ^
It is .6 because it is either him, or maybe not him, or maybe him from another part of the multiverse.
Thanks for the cutting edge analysis, MSU. Did any of my tax dollars go to supporting this work?
November 6th, 2023 at 5:29 PM ^
"If I take an identical photo, it would be one."
Yes
November 6th, 2023 at 3:51 PM ^
So CMU played at Michigan State. How does Michigan State have no idea who was on their field? Would they not have issued all of the passes for both teams?
How has CMU not gone down the list and said Joe Smith, you had a sideline pass: where you there?
November 6th, 2023 at 4:07 PM ^
"Our top story tonight: Generalissimo Francisco Franco is still dead..."
November 6th, 2023 at 4:09 PM ^
In other news trust in the media collapses to historic lows per Gallup.
November 6th, 2023 at 1:59 PM ^
I refuse to click on an ESPN story.
November 6th, 2023 at 2:00 PM ^
Central got some 'splainin to do.
The FOIA requests for phones/emails from that week will be highly interesting. Somebody knows who that person is (lol) and hasn't stepped forward. Could it be McElwain?
November 6th, 2023 at 2:04 PM ^
I cannot recall a single recent story in CFB or even the general media where so much “known” information and evidence has taken this long to confirm or be released.
November 6th, 2023 at 2:10 PM ^
I'm wondering why he isn't under a microscope more than he has been for all of this.
Isn't the whole point of this farce is that the HC is responsible for everything? And can't plead ignorance?
November 6th, 2023 at 2:20 PM ^
McElwain is still under investigation by the Woods Hole Oceanographic Institute.
November 6th, 2023 at 2:30 PM ^
That's wood hole to you, bub. NTTAWWT. Or is there? Questions of consent, I would imagine.
November 6th, 2023 at 3:22 PM ^
S.D. Jones is killing it today. I am envious!
If I ever get to Bolivia I'm coming back as M.F. Jones, unless of course if it's already taken.
November 6th, 2023 at 4:23 PM ^
Thank you DH, you're too kind. Would love to have you in the Jones family.
November 6th, 2023 at 2:24 PM ^
Not sure, but I would bet every dollar I have that no one on Michigan's staff told Stallions to go to Central vs MSU
November 6th, 2023 at 2:45 PM ^
Sounds like you're pretty sure .
November 6th, 2023 at 4:15 PM ^
Or he is broke.
November 6th, 2023 at 3:46 PM ^
It seems like freelance work to me. There is dumb, but turning up on the CMU sideline is village idiot level dumb!
November 6th, 2023 at 2:27 PM ^
McElwain could be in trouble if he wrote Connor a check for services rendered.
November 6th, 2023 at 2:35 PM ^
Agreed and raised this elsewhere - if he did invite him - he's probably technically on CMU's staff. Does that make it a violation if he is a dual, though temporary, staff member? Certainly isn't contemplated, but certainly makes Michigan look less bad (would never be cleared internally) and frankly make it not even a violation in the first place.
November 6th, 2023 at 2:54 PM ^
I can't figure out how this would not also be a violation for cmu? It's pretty obvious that if that is Stallions, someone on CMU allowed him access. I can see how that benefited cmu, but how does that help uofm? We didn't play cmu this year