PlaySlot learns object-centric latent actions from unlabeled video and uses them for controllable future-frame prediction and sample-efficient robot behavior learning.
In contrast to the object-centric representations employed by PlaySlot, LAPO relies on feature maps output by a convolutional encoder
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
PlaySlot: Learning Inverse Latent Dynamics for Controllable Object-Centric Video Prediction and Planning
PlaySlot learns object-centric latent actions from unlabeled video and uses them for controllable future-frame prediction and sample-efficient robot behavior learning.