Hacker News new | ask | show | jobs
by astroflask 2163 days ago
Earlier this year I made a video[0] where I took two images which are frequently posted on reddit that show the movement of martian sand over a day and interpolated (with python and opencv) the movement into a smooth video. I'm currently working on restoring some Apollo footage, but one of the pending projects is taking the Curiosity descent sequence (at 4 frames per second) and give it a shot at some nice quality, XXI century interpolation (we have some very good algorithms now, or at least better workarounds the artifacts).

[0] https://youtu.be/k7pfdFMVj-o
1 comments
garaetjjte 2163 days ago
>give it a shot at some nice quality, XXI century interpolation

I think it has been done? https://www.youtube.com/watch?v=tMntZ6_R78Q

link
astroflask 2163 days ago
The creator on that channel simply uses a DAIN application without pre or post processing the images (either the base images or the optical flow maps).

I take a different approach and I can't say I haven't found issues (I have, many, I'm working on them).

My goal is to reach a level where there are almost none artifacts (there are a couple in that video at the start, when the shield is released and when there are many large movements of the camera due to Curiosity dangling off the parachute). As I said, I'm still working on it, and in between my other side projects, work and quarantine I've been on a less-than-ideal situation.

In time, if I pull it off, I'll post it and I guess you'll see it.

link