This video really struck me. For the imagination and the technique used.

Until now, I hadn't seen underwater videos generated by A.I. that were not just convincing, but even interesting. Bravo!!

Edit: I'm sad for those lens underwater 😛

When fashion holds its breath... 🌊 Several meters deep,...

When fashion holds its breath... 🌊 Several meters deep, fashion claimed the ocean floor. Submerged catwalks between corals and shipwrecks, silhouettes moving through the silence of the abyss, under

How about using AI for fish identification with a HUD underwater?

I think there are projects for fish ID.. Google does a pretty decent job of it already, but one issue of UW is that we don't have a good way to connect to a remote server (EM waves don't travel well through water). We can run a pre-trained AI "in-mask". We'd would need a fairly powerful computer.. maybe Arduino will not cut it - perhaps need something like a Raspberry Pi. Plus you need to power it, I guess battery life would suck. I think it's a great project. It should come first to those UW rovers, which are larger and already have cameras..

Cheers

Ajay

We've looked into running object detect and identification models at the edge (on-device with smaller hardware) and the biggest issue ends up being compute and storage. Most models today run on GPUs, thus the massive data centers and the associated run-up in nVidia stock! We've been able to run some of the smaller models on-device, but without cloud connectivity, something like a species identification model may be a bit too far for current hardware. You may get something that will identify at a larger level (whale, shark, porpoise, etc), but granular ID could be tough.