Great work, came across this just yesterday, lots of citations already! I wish y'all and all the folks in this field good luck to continually come up with cool stuff!
Take a drink every time he says "barf". Also, smart work. I love work that utilizes a coarse-to-fine way of thinking about solving problems. ::) Also the previous work's results look like barf. The algorithm just barfs all over it - barf in, barf out, you could say. But now, I can't wait to barf all over my friends with my phone.
hello amazing work, i have a question regarding the implementation of barf, if we want to use the model on our own dataset do we need to specify the focal length of the camera?
Is it sensible to merge this with Thomas Neff's work around DoNeRF to provide real-time reconstruction while also taking advantage of your not knowing camera pose? Great work!
![[ICCV 2021] Pixel-Perfect Structure-from-Motion with Featuremetric Refinement](http://i.ytimg.com/vi/2HuCMuraFk0/mqdefault.jpg)
![[ICCV 2021] Pixel-Perfect Structure-from-Motion with Featuremetric Refinement](/img/tr.png)
Great work and sense of humor :)
Great work, came across this just yesterday, lots of citations already! I wish y'all and all the folks in this field good luck to continually come up with cool stuff!
Amazing video and work! Thank you.
Take a drink every time he says "barf".
Also, smart work. I love work that utilizes a coarse-to-fine way of thinking about solving problems. ::)
Also the previous work's results look like barf. The algorithm just barfs all over it - barf in, barf out, you could say. But now, I can't wait to barf all over my friends with my phone.
Great work!
Awesome !!!!
If NERF was awesome per-se... You make it even more more more awesome !!!!
hello amazing work, i have a question regarding the implementation of barf, if we want to use the model on our own dataset do we need to specify the focal length of the camera?
Noice 👍
I'm surprised that none of the authors names are Tony Stark or Quentin Beck
Is it sensible to merge this with Thomas Neff's work around DoNeRF to provide real-time reconstruction while also taking advantage of your not knowing camera pose? Great work!
Thanks! Yes, I do think there are different ways of utilizing ground-truth depth to increase sampling efficiency for coarse-to-fine registration 🙂
I like how the smilie is just part of the official paper name, lol
I like the meme picture used in this.😆
Nice work! I think 4:36 it's f(x + c) instead of f(x) + c
Yes great catch, thanks for pointing it out! That's certainly a typo 😅