Thank you again UbiOps for sponsoring this video. I hope you guys enjoyed it, and I am sure you will enjoy what UbiOps can offer! Click here to sign up for a free account at UbiOps: ubiops.com/product/
The title of your video is misleading. You did not achieve slow motion with AI only, unlike motion interpolation plugins, you require a special camera that shoot at the high frame rate you need, even though at a lower spatial resolution.
Super interesting, thanks for sharing! If I understand well, the event-based camera films or registers only the pixels that change, right? So that the image can be reconstructed later?
Exactly! And it records at a much higher frame rate (fps) since it has much less information and of less quality to process (1280*720) I believe. Then, basically, using the information in the frames they can interpolate the events and reconstruct all the pixels of each event that are in-between the frames!
Sorry it was misleading. This was not my intention! I wanted to explain the approach introduced in the paper and not the code itself, which I believe is what you were looking for. Sorry for not delivering what you expected! I don’t usually go over the code as I personally prefer the ideas and mathematical/algorithmic concept itself, which I believe is more interesting (personal opinion of course, I know that the code is what ultimately make it work haha). Anyways, hope the code is clear and you manage to use it! Feel free to tune in for paper and techniques explanation when you need those!
I agree that I could add that to my videos in the future. I should run some pills to gauge the interest regarding the code and implementation + maybe how I ran it (if I had to edit anything or not to make it work)! That may be really interesting. Thank you for the suggestion!
![Better Than DAIN? NEW BEST Tool for Boosting Video's FPS with AI [RIFE/Flowframes]](http://i.ytimg.com/vi/60DX2T3zyVo/mqdefault.jpg)
Thank you again UbiOps for sponsoring this video. I hope you guys enjoyed it, and I am sure you will enjoy what UbiOps can offer!
Click here to sign up for a free account at UbiOps: ubiops.com/product/
The title of your video is misleading. You did not achieve slow motion with AI only, unlike motion interpolation plugins, you require a special camera that shoot at the high frame rate you need, even though at a lower spatial resolution.
Very interesting, never heard of such event cameras! Very good choice of topic and good presentation!
Thank you, Tristan! Glad you think so. :)
Super interesting, thanks for sharing! If I understand well, the event-based camera films or registers only the pixels that change, right? So that the image can be reconstructed later?
Exactly! And it records at a much higher frame rate (fps) since it has much less information and of less quality to process (1280*720) I believe. Then, basically, using the information in the frames they can interpolate the events and reconstruct all the pixels of each event that are in-between the frames!
@@WhatsAI So cool... thanks a lot for the precision !
My pleasure!
Amazing
you did NOT explain HOW!
your title is misleading, i went to GitHub, disliked your video and install on my own.
thanks for nothing i guess
Sorry it was misleading. This was not my intention! I wanted to explain the approach introduced in the paper and not the code itself, which I believe is what you were looking for. Sorry for not delivering what you expected! I don’t usually go over the code as I personally prefer the ideas and mathematical/algorithmic concept itself, which I believe is more interesting (personal opinion of course, I know that the code is what ultimately make it work haha).
Anyways, hope the code is clear and you manage to use it! Feel free to tune in for paper and techniques explanation when you need those!
@@WhatsAI i mean you could have showed how to compile the code from GitHub and show how to do a test run.
no worries, thank you for clarifying
I agree that I could add that to my videos in the future. I should run some pills to gauge the interest regarding the code and implementation + maybe how I ran it (if I had to edit anything or not to make it work)! That may be really interesting. Thank you for the suggestion!