search

facebookresearch / DeepFovea

Neural Reconstruction for Foveated Rendering and Video Compression using Learned Statistics of Natural Videos
Other
387 stars 55 forks source link

Any ETA when the trained model will be published? #1

Closed reneschulte closed 4 years ago

reneschulte commented 4 years ago

Intentionally left blank. 😉

mmehas commented 4 years ago

Hi Rene, thanks for your question. We are going through some internal approval processes to publish the model and we don't have a specific ETA that we can share, sorry. It shouldn't take too long, though.

reneschulte commented 4 years ago

Why is this closed? Can't we wait until it is actually released?

mmehas commented 4 years ago

I was assuming that since we provided the answer, the issue could be closed. If that would help, let's reopen it. We will post an update here when we publish the model.

reneschulte commented 4 years ago

Thank you @mmehas Really appreciate it. For me it's a great way to make sure I don't miss it and get a notification in my mailbox.

CianNoonan commented 4 years ago

I appreciate that Christmas has probably derailed this but any update on the approval for release here? I'm really interested in seeing the full dataset at work

mmehas commented 4 years ago

Hi @CianNoonan, we appreciate your interest, and sorry for the delay. The process is almost finished though and we should be able to publish the working model in the upcoming weeks.

DenisSergeevitch commented 4 years ago

Hi, any news about the trained model? :)

mmehas commented 4 years ago

@DenisSergeevitch, at the moment we are done with the technical details and preparing a short blogpost\documentation to accompany the release.

DenisSergeevitch commented 4 years ago

@mmehas looking forward to try it, you are amazing :)

jonnyawsom3 commented 4 years ago

Just found this via youtube (presumably like many others) and I'm quite interested by it's video compression capabilities.

Glad to hear there's activity and hopefully one day someone will be able to whip up a small GUI application for it. (So compressing a video would be "select input video, compression percentage, output name" and then "select video, output name" for... Extraction? I'm not quite sure what to call it... Interpolation may be the right word)

Hope to hear more soon!

kaplanyan commented 4 years ago

Today, we are excited to release the full DeepFovea demo, so the researchers can experience the neural foveated rendering and reproduce the results of our DeepFovea paper either on a screen or in a headset. Thanks to the novel generative AI algorithm, DeepFovea sets a new milestone in foveated rendering with only 10% of pixels needed to reconstruct the full image. Please check out the new instructions on how to run the demo. If you are interested to learn more about how DeepFovea was designed, you can check out this new post as well https://research.fb.com/blog/2020/05/deepfovea-ar-vr-rendering-inspired-by-human-vision/

Best, The DeepFovea team

reneschulte commented 4 years ago

Awesome. Thanks @kaplanyan 👏