This is officially shitposting at this point. Like what did one expect to see from these results? Is this even interesting in any way?
Yeah this is kind of dumb. “Here’s a blurry picture of John Oliver”.
I needed to align each frame so John's face (if present) would stay in place.
... That pretty much ruins the picture. The only interesting thing I took away from the picture was that John Oliver was almost always in the same place - now it turns out that's just a choice you made. Of course you're going to get a blurry picture of John Oliver if you specifically keep him in the same place.
These posts are so stupid. Can we actually post interesting data? I'm sorry but blurry pics of people(s) just isn't that interesting.
Ooooh so that's who it is!
I thought it was british Milhouse... well that or a puffin in a suit...
I'm still deciding tbh.
Hes not rat faced, he looks like a near sighted parrot who works at a bank
I downloaded 36 videos (a little over 9 hours of TV) from Last Week Tonight's YouTube channel with youtube-dl. Extracted the frames into jpegs with ffmpeg. Used OpenCV and python to automatically align the frames so John's face would always be in the same place when it was present. Added up all the pixel values and finally normalized into RGB so the darkest pixel value would be 0 and the brightest 255, for better contrast.
Here's a visualization where I'm showing different subsets of the data in each GIF frame:
This is impressive but your result is the most boring thing I've seen on the internet
Whoa! It looks like John Oliver! In a suit! At a desk! With a square image floating near his head! It looks like Last Week Tonight!
Here's an example of the kind of content that is in that million frames: https://imgur.com/a/jLW27
It's not just a handful of pictures of John Oliver sitting behind a desk. Kathy Griffin is in there, and so are Stephen Hawking and the sloth.
It wasn't obvious to me at all whether you'd even get a recognizable image after blending a million frames. I needed to align each frame so John's face (if present) would stay in place. Otherwise you get an effect like this: (just one video used for this early result)
It kind of sounds like we can finally get rid of this voting system and let you choose what's on the frontpage.
Nope, 9 hours are 32400 seconds. That times 30 is 972000 frames. He says little over 9 hours so it could definitely be 1 million frames
I thought it was Steve Mnuchin
I did mention that in the first comment. FWIW, here's the full non-aligned version:
Suprised that Trumps face doesn't show up.
He's a rat faced bastard with creepy spider hands.
Sounds like you've given yourself a challenge. Can you rise to the occasion?
You can kinda make out an orange smudge in the box, top left
I guess I understand where you're coming from, but it feels fairly obvious that if you take a large sample of time from a show that focuses around a single host who for the majority of most shows is positioned in the same place, you're gonna get an image that is essentially that person but blurred.
I don't deny that this took a lot of work, aligning the frames especially, but a) I don't believe that makes it worthwhile and b) doesn't aligning the frames almost defeat the point, since that doesn't reflect the actual result of blending said frames?
That's fair enough. I think the journey was more interesting than the destination, and I wanna share a bit of that. https://imgur.com/a/M8jqW
These are the individual videos before final merge. It's still just a John Oliver and some noise, don't get me wrong, but you can see the variety and there's a little more detail. True fans will recognize every episode :)
That's President Orange Smudge, mister
I gathered the material way back in March of 2016. I would estimate I have roughly 1/7 of all the episodes that have aired to date.