Johannes Kopf (@JPKopf) profile picture

Vision / Graphics / ML Researcher @ Facebook Reality Labs

Seattle
Joined June 2014
Photo by Johannes Kopf (@JPKopf)Photo by Johannes Kopf (@JPKopf)Photo by Johannes Kopf (@JPKopf)Photo by Johannes Kopf (@JPKopf)Photo by Johannes Kopf (@JPKopf)
Also, look at those reflections!!! 😊
Bullet-time on any video!
1
2
Bullet-time on any video!
SUPER excited to share our work Dynamic View Synthesis from Dynamic Monocular Video! Check out our fun teaser video 😎! Paper: arxiv.org/abs/2105.06468 Project: free-view-video.github.io/ w/ @gaochen315, @ayush29feb, @JPKopf
4
1
27
No one can stop you from trading Dogecoin!
1
6
People who say $GME is overvalued don't understand the value proposition. It's no longer a stock, it's a store of value. It's the new digital gold. And for now it has only 3% of the $BTC market cap, still plenty of room to grow. My price target is over $9,000
195
607
320
6,075
Suckers like me only buy bitcoin at the all-time high... 🤦🏻‍♂️
3
Johannes Kopf retweeted
Space-Time Neural Irradiance Fields (video-nerf.github.io/) simply use time as an additional input. Carefully selected losses are needed to successfully train this method to render free-viewpoint videos (from RGBD data!).
1
3
1
18
You can try this algorithm *right now*, on your own phone, with your own photos. It's integrated in your Facebook app! :)
This quoted post is unavailable.
1
16
Thank you for the beautiful coverage of our paper, Károly! Find all the details about the work here: facebookresearch.github.io/o…
This quoted post is unavailable.
13
You’re revealing our best secrets...
Replying to @jbhuang0604
How can I get my paper on the top? Understand that arXiv is a "stack" data structure (first in last out, FILO)! Submit your paper RIGHT BEFORE THE DEADLINE will make your paper appear on the top! Probably the easiest way to improve your paper visibility.
1
2
Wow, this is really cool! I didn’t know it’s a thing and people create so many of these. We’re kinda aiming at doing this in 3D -> So you can change the viewpoint interacively while it’s playing.
1
7
Robust-CVD works great for long videos as well. Such a fun tool to use! arxiv.org/abs/2012.05901
4
1
53
Hey! Algorithmically it’s relatively straightforward, and these methods will work well for 360. But there’s quite a bit of engineering involved. We’re looking into doing some of that work, though, because 360 *3D* videos will look absolutely amazing in VR!
Replying to @JPKopf @jbhuang0604
How are you? I have a question based in my singulars dreams, did you make a test on a 360 degree video, if not, it is possiblr to archirve a good result? Greetings from.argentina
1
3
Our latest work on making Consistent Video Depth more ROBUST. Works great for casual phone videos that are really difficult for previous methods. Another great collaboration with @jastarex and @jbhuang0604. arXiv: arxiv.org/abs/2012.05901 Project: robust-cvd.github.io/
4
54
7
271
We test the robustness by applying the method to ALL videos of existing video datasets. So, no cherry picking! :) Here’s all 90 DAVIS videos.
1
1
1
14
Now, just need to combine this with our recent space-time free-viewpoint video tech. Stay tuned!
Our awesome new paper on *free-viewpoint* video from cell phone capture. Space-time neural rendering with the powerful @wenqi_xian, @jbhuang0604, and @_ChangilKim! Paper: arxiv.org/abs/2011.12950 Website: video-nerf.github.io/
2
11
Killer results, Aleks. This looks even improved compared to the last time I’ve seen it?
Excited to show off our new project on single-image cinemagraphs. Our method automatically turns a _single image_ into a seamlessly looping video! Website: eulerian.cs.washington.edu Video: youtube.com/watch?v=4zKliOMi… w/ Brian Curless, Steve Seitz, Rick Szeliski More in thread! [1/5]
1
6
Doesn’t look like a failure case to me :D
Replying to @holynski_
And also some pretty cool failure cases -- if a class of objects aren't seen during training, but share similar textural properties to fluids... [6/5]
1
Seems it’s time to get my bitcoin ticker out of the closet again.
1
This looks GREAT, Oliver. Awesome work!
Check out our latest work where we perform space-time view synthesis of dynamic sequences from monocular video! w/ @zl548 @Jimantha @simon_niklaus. cs.cornell.edu/~zl548/NSFF/ I'm really excited about this work and there are lots of cool things to unpack, some favorites: [1/4]
3
Honestly, this was a five-minutes before the deadline change, without _too_ much thought into it :)
Replying to @jon_barron @JPKopf
Interesting discussion! I think radiance relative to irradiance is more appropriate. Human eyes (& cameras) which have a lens, don't have direct access to the irradiances but radiances. And irradiance-meter varies w. distances, which the demo (very impressive!) doesn't show.