SlotDiT: Diffusion Transformer uses object-centric slots for video generation

Researchers have introduced SlotDiT, a novel Diffusion Transformer (DiT) model designed for video generation and robotic applications. This model operates within a slot-based latent space, which decomposes scenes into object-centric representations. SlotDiT leverages these structured latents to predict future scene dynamics based on language instructions and observed context. Experiments indicate that SlotDiT achieves competitive video generation quality and enhances task completion rates in robotics, offering a more computationally efficient alternative to VAE-based methods. AI

IMPACT Introduces a novel object-centric representation for diffusion models, potentially improving robotic control and video generation efficiency.

RANK_REASON The item is a research paper detailing a new model architecture and its application. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Diffusion Transformer
• Gotit.pub
• Hugging Face
• ScienceCast
• SlotDiT
• variational auto-encoder

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →