Beautiful Jaclyn Smith, Kelly Garrett in Charlie's Angels 1976-1981, the longest serving Angel.
More from All
How can we use language supervision to learn better visual representations for robotics?
Introducing Voltron: Language-Driven Representation Learning for Robotics!
Paper: https://t.co/gIsRPtSjKz
Models: https://t.co/NOB3cpATYG
Evaluation: https://t.co/aOzQu95J8z
🧵👇(1 / 12)
Videos of humans performing everyday tasks (Something-Something-v2, Ego4D) offer a rich and diverse resource for learning representations for robotic manipulation.
Yet, an underused part of these datasets are the rich, natural language annotations accompanying each video. (2/12)
The Voltron framework offers a simple way to use language supervision to shape representation learning, building off of prior work in representations for robotics like MVP (https://t.co/Pb0mk9hb4i) and R3M (https://t.co/o2Fkc3fP0e).
The secret is *balance* (3/12)
Starting with a masked autoencoder over frames from these video clips, make a choice:
1) Condition on language and improve our ability to reconstruct the scene.
2) Generate language given the visual representation and improve our ability to describe what's happening. (4/12)
By trading off *conditioning* and *generation* we show that we can learn 1) better representations than prior methods, and 2) explicitly shape the balance of low and high-level features captured.
Why is the ability to shape this balance important? (5/12)
Introducing Voltron: Language-Driven Representation Learning for Robotics!
Paper: https://t.co/gIsRPtSjKz
Models: https://t.co/NOB3cpATYG
Evaluation: https://t.co/aOzQu95J8z
🧵👇(1 / 12)
Videos of humans performing everyday tasks (Something-Something-v2, Ego4D) offer a rich and diverse resource for learning representations for robotic manipulation.
Yet, an underused part of these datasets are the rich, natural language annotations accompanying each video. (2/12)
The Voltron framework offers a simple way to use language supervision to shape representation learning, building off of prior work in representations for robotics like MVP (https://t.co/Pb0mk9hb4i) and R3M (https://t.co/o2Fkc3fP0e).
The secret is *balance* (3/12)
Starting with a masked autoencoder over frames from these video clips, make a choice:
1) Condition on language and improve our ability to reconstruct the scene.
2) Generate language given the visual representation and improve our ability to describe what's happening. (4/12)
By trading off *conditioning* and *generation* we show that we can learn 1) better representations than prior methods, and 2) explicitly shape the balance of low and high-level features captured.
Why is the ability to shape this balance important? (5/12)
You May Also Like
1/Politics thread time.
To me, the most important aspect of the 2018 midterms wasn't even about partisan control, but about democracy and voting rights. That's the real battle.
2/The good news: It's now an issue that everyone's talking about, and that everyone cares about.
3/More good news: Florida's proposition to give felons voting rights won. But it didn't just win - it won with substantial support from Republican voters.
That suggests there is still SOME grassroots support for democracy that transcends
4/Yet more good news: Michigan made it easier to vote. Again, by plebiscite, showing broad support for voting rights as an
5/OK, now the bad news.
We seem to have accepted electoral dysfunction in Florida as a permanent thing. The 2000 election has never really
To me, the most important aspect of the 2018 midterms wasn't even about partisan control, but about democracy and voting rights. That's the real battle.
2/The good news: It's now an issue that everyone's talking about, and that everyone cares about.
3/More good news: Florida's proposition to give felons voting rights won. But it didn't just win - it won with substantial support from Republican voters.
That suggests there is still SOME grassroots support for democracy that transcends
4/Yet more good news: Michigan made it easier to vote. Again, by plebiscite, showing broad support for voting rights as an
5/OK, now the bad news.
We seem to have accepted electoral dysfunction in Florida as a permanent thing. The 2000 election has never really
Bad ballot design led to a lot of undervotes for Bill Nelson in Broward Co., possibly even enough to cost him his Senate seat. They do appear to be real undervotes, though, instead of tabulation errors. He doesn't really seem to have a path to victory. https://t.co/utUhY2KTaR
— Nate Silver (@NateSilver538) November 16, 2018
A list of cool websites you might now know about
A thread 🧵
1) Learn Anything - Search tools for knowledge discovery that helps you understand any topic through the most efficient
2) Grad Speeches - Discover the best commencement speeches.
This website is made by me
3) What does the Internet Think - Find out what the internet thinks about anything
4) https://t.co/vuhT6jVItx - Send notes that will self-destruct after being read.
A thread 🧵
1) Learn Anything - Search tools for knowledge discovery that helps you understand any topic through the most efficient
2) Grad Speeches - Discover the best commencement speeches.
This website is made by me
3) What does the Internet Think - Find out what the internet thinks about anything
4) https://t.co/vuhT6jVItx - Send notes that will self-destruct after being read.
