I'm considering doing some ML stuff for a mobile DJ app.. like beat/bpm detection, instrument / vocal separation etc. Have you seen anything recent that might be efficient enough to run on a mobile device and process a track in a reasonable amount of time ( less than song length ) ?

Made a simple and free pushup counter app that uses the depth sensor on the iPhone to count push ups. https://PushupCounterApp.com The existing apps I tried were subscription based / unreliable / overcomplicated. It's a WIP since I hacked it together over a few days.

For the motion part, there is already an extension to gaussian splatting for moving stuff: https://dynamic3dgaussians.github.io/ You can also just string a bunch of them together to create an animation: https://twitter.com/8Infinite8/status/1699460316529090568 For relighting, there are lots of NeRF variants that do this -- it should be possible to optimize material parameters for the splats.

Here is a similar idea using 15 iPhones + NeRF rendering to make a bullet time video https://labs.laan.com/blogs/slow-motion-bullet-time-with-ner...

Link 1: “My cousin was getting married last month, and we were doing a family photo shoot for the occasion”

Link 2: “ For my cousin’s wedding I did not make a photo booth but a video booth.”

That’s n = 2 for cousins getting married inspiring bullet cam implementations.

I'm eagear for the sequel, divorce bullet time, it may not be appreciated by any of the subject but is hard to argue that mid-air tears in bullet time would be quite a sight to see.

If you have a compatible mac there are several free apps like this one: https://apps.apple.com/us/app/3d-scanner-app/id1419913995

I had a two letter name which got hacked. I called in a favor from a friend of a friend at instagram/FB and got it back.. then it happened again and I didn't want to ask the favor again. IIRC they did not yet have 2FA even though I asked for it ( I was assuming it would happen again and it did. )

I made a simple static site that processes user images and videos with WASM and it would be great for users to be certain that their data will not leave the browser. It seems possible to define some list of resources I plan on loading, and anything else should not be allowed.

If you have an extra smartphone, you can use some of the AR SDKs out there. ARKit has the ability to relocalize itself to a stored map ( e.g. your apartment ) The 6D.ai SDK can even do meshing for obstacle avoidance.

I was just going to suggest piping this into a pix2pix / CycleGAN trained on face edges for some creepy looking results.

One option is to submit the app under a new dev account with a few tweaked variables ( name, colors, screenshots ) I feel like some of these rejections depend on the reviewer you get that day; and once an app is flagged, it's hard to recover.

This is a really bad idea. These platforms like Apple and Google control large online ecosystems, and they're known to permaban people across the board when they step out of line. There are dozens of stories about Google basically terminating all relationships with someone (including their Gmail, Drive content, etc.) over Play Store dev account hijinks.

I presume the author of the article has an iPhone. If he intends to be able to use an iPhone, screwing around with making Apple accounts that pretend to be different people to sneak an app past the reviews is not a good decision. The developer could find themselves unable to use any Apple account or service.