RandomPrototypes

[RandomPrototypes @RandomPrototyp1]

わたしは「SASUKE VR」の世界を揃えたよ！！
楽しかっかです！
#ひろげようキミごと世界を
#MetaQuest
@MetaQuestJapan
https://t.co/DI5ak3zqW3

[RandomPrototypes @RandomPrototyp1]

@Raspberry_Pi Event cameras are awesome but currently too expensive. Hopefully this module gets a reasonable price but I doubt it after seeing "request a quote".

[RandomPrototypes @RandomPrototyp1]

@SkarredGhost Unity has a class called TextureConverter that allows to convert a WebcamTexture to a Sentis Tensor directly. I'm not sure how it is implemented internally but I can hope it's well optimized.
https://t.co/Sv5M3xlKSv

[SASUKE【TBS公式】 @sasuke_tbs]

すごすぎる！
これがミスターSASUKEだ！

#山田勝己 さん
3ミニッツプッシュアップ優勝🏆

これは感動🥹🥹🥹🥹
本当にすごかった！
おめでとうございます🎉

選手の皆さんも挑戦本当に素晴らしかったです😭

[RandomPrototypes @RandomPrototyp1]

@SkarredGhost with camera access, quite easily.
without camera access, it requires an annoying calibration process like for mixed reality capture and cannot live-recalibrate to compensate for drift while moving.

[RandomPrototypes @RandomPrototyp1]

@cixliv @VibeFitnessGame If you try to make it work on older devices, you'll probably get trouble with thermal throttling unless the AI model is very light. It may be worth to add external cooling in the dock to avoid losing ~75% of the AI power on older devices.

[RandomPrototypes @RandomPrototyp1]

@SkarredGhost @UploadVR Open-source / simple to decode protocol or locked to pico ecosystem only?
Would be great if they add optional external cameras for higher accuracy mode or for use for other purpose than VR.

[RandomPrototypes @RandomPrototyp1]

@XolotlVtuber @NimaZeighami I tried on quest 2 ( scrcpy -> OpenCV -> tag detection) for AR but the resolution and distortion were too bad to detect the AR tags except when perfectly centered.
I don't have quest 3 to test it.

Another problem is that if we render over the tag, it cannot be detected anymore

[RandomPrototypes @RandomPrototyp1]

@CixLiv they should at least implement AR tags detection

[RandomPrototypes @RandomPrototyp1]

@NimaZeighami oh, ok. I thought it was like video streaming to sidequest or scrcpy.

[RandomPrototypes @RandomPrototyp1]

@CezaryXR @Lunayian The leap-motion sensor, for example, was not a depth-camera but an IR illuminator with 2 IR cameras. Close range objects (hands,...) became bright while far range was less illuminated. This helped to get high quality segmentation and do precise depth estimation by stereo.

[RandomPrototypes @RandomPrototyp1]

@CezaryXR @Lunayian Depth sensors generally have a lower resolution than cameras and are not good for thin structures (fingers,...) but are good for uniform surfaces at which stereo-cameras are bad.
It would be interesting to know the technology of the depth cam (structured light, tof,...).

[RandomPrototypes @RandomPrototyp1]

@Fallacy_Vortex @MetaQuestVR you can disable the guardian in the settings on quest 2 also, but the passthrough view is in grayscale and low quality.

[RandomPrototypes @RandomPrototyp1]

@MetaQuestVR But the quest loses 90% of its AR potential by not providing camera access to the developers 😭
Interaction with moving object cannot be recognized. We cannot detect when a plate is washed or the location covered by the vacuum cleaner (could spawn some enemies to vacuum),...

[RandomPrototypes @RandomPrototyp1]

Interesting full-body-tracking research I saw at ICCV23 from Eth Zurich and Meta reality labs.
They use electromagnetic sensors for full-body-tracking without occlusion problem, and they refine it from a single rgb camera
#VR #MetaQuest #fullbodytracking

https://t.co/YIk3qg5Rj4

[RandomPrototypes @RandomPrototyp1]

@VoodooDE_Gaming @MetaQuestVR If you can test my app, I'd love to see how it looks with color passthrough
https://t.co/LAusMmgku4