Sometimes I learn a piece of new information that makes me say, "That's enough. I am done learning for today. I don't want to read or think anymore."
https://t.co/1s8F0yKBK9
\u201cDrew certainly isn\u2019t a newcomer to the tougher moments of this country\u2019s history. At the very beginning of his career he was one of only four photographers in the kitchen of the Los Angeles hotel when Bobby Kennedy was shot and killed.\u201d\U0001f914\U0001f9d0 https://t.co/JQ6M9qgq0N
— Alfredo Gomes (@mindgomes) May 4, 2021
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)