
I like vision models @pokerpanel
Austin, TX
Joined November 2020
- Tweets5.4K
- Following89
- Followers363
- Likes887
hi. i am that college student. i also need an internship for my degree!
some technical attributes to correct:
video is not iproxy, i tried that, it failed. i use Media over QUIC, a modern data transfer system that uses UDP and was pioneered for video transport specifically and used for cloud gaming.
i didnt manually label thousands of frames, i did do hundreds initially, but ai is pretty good now at labeling data now. i do also have a really strong synthetic dataset.
the reason i use iphones is because a non P2P wireless video system was a requirement (apartment internet had strong security settings). OBS does not support this, so i built my own broadcast software along with Poker Panel.
RFID tables do not cost 2K minimum, they are much more. RFID card durability is also known to be weak.
youtube.com/watch?v=YAuiqCIc…
No one is more hyped for astra than ml engineers working with real world physical datasets
Replying to @SmokeAwayyy
The real world implications of high tps rendering/asset generation will easily substitute ml datasets regarding vision models and modern robotics.
88
This is the first realistic solution for card rooms that want to livestream. If you use anything else you will have a worse time.
Introducing Card Room Tier.
Card Room Tier will focus intensely on branding, game integrity, and incorporating niche broadcast related solutions for users.
Card Room Tier will give developers an api to directly pull stats in realtime.
Pricing is currently 299/mo.
95
The “race to zero” argument is wrong.
Replying to @Dr_Singularity
As tokens get smarter tokens will get better at allocating other tokens. The race to zero argument works if you assume # of tasks completed remains the same.
63














