(1/5) Here’s a thread about some fun technology we (with @_Quazi_ and @anorangeduck), are working on at La Forge (our research lab inside Ubisoft Montreal).
The TL;DR is that we trained a self-balancing physics-based character to follow interactive motion capture with DeepRL.pic.twitter.com/U7JEjUiuZ1
-
-
(4/5) If we just give rewards for going at the desired speed, he forgets the motion-captured style and looks a bit silly.
pic.twitter.com/OKbVPIsTjXPrikaži ovu nit -
(5/5) Next week we’ll present this method at
#SIGGRAPHAsia2019. The paper is titled “DReCon: Data-Driven Responsive Control of Physics-Based Characters”. My colleague Kevin Bergamin wrote a nice blog post with more details:https://montreal.ubisoft.com/en/drecon-data-driven-responsive-control-of-physics-based-characters/ …Prikaži ovu nit
Kraj razgovora
Novi razgovor -
-
-
I imagine the poor artificial agent waking up in the morning and saying: "Today it's going to be a good day of training". Hang in there buddy.
- Još 1 odgovor
Novi razgovor -
-
-
Throwing pixel rocks at a pixel character to try to make it trip for the sake of testing your physics engine looks like such a fun job! Maybe simply because I felt the "sometimes you just get to many cubes in the face"...
-
Mondays.
Kraj razgovora
Novi razgovor -
-
-
Man, I hate it when that happens
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
- Kraj razgovora
Novi razgovor -
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.