5 ms·
It's very cool, but the demo videos suggest that the tracking isn't ideal yet. Which susprises me, because we only have what, 50+ years of R&D with tracking mo
by brianstorms 11y ago
It's very cool, but the demo videos suggest that the tracking isn't ideal yet. Which susprises me, because we only have what, 50+ years of R&D with tracking moving targets. As in, if you can figure how how fast the target is moving across the field of view, you should be able to predict where it is going to be in the next frame, PLUS, take into account the latency of the Soloshot motor, and calculate that all together in real-time such that the target stays as close as possible to the center of the field of view. Right now, based on the demo videos, it appears that either the CPU is slow, or the Soloshot motor is slow, or both, with the result being that the target gets ahead of the Soloshot and gets to the edge of the field of view regularly, making for less than optimal viewing afterward.
Seems like a totally, well, 99% fixable problem, and given Soloshot's been out for a while I'm surprised the demos aren't more stellar.
Still, it's impressive and I'm surprised GoPro hasn't acquired the company already.
- IanDrake 11y agoIn practice, tracking has never been a problem for me. Have you seen the demos of it tracking quad copters?
- bitJericho 11y agoThe skier video on the website is just plain bad. The camera should zoom out slowly and smoothly, instead it was jerky.
- IanDrake 11y agoYeah, that one is bad. If I had to guess, I'd say that the further away you are, the more the camera is zoomed in, the more jagged the panning will seem, as the distance in one degree of arc at 100ft is much less than 1000ft. I imagine this could have been easily taken out in post production.
- emcrazyone 11y agowas having similar thoughts. I was also wondering why the soloshot had to track a device you wear instead or in addition to something like openCV that can be trained on your face or physique.