Skip to content
Agientry
← Back to Videos

COSMI: Compositional Human-Object Interaction Dataset and Diffusion Model

Loading video
Loading video

COSMI composes existing single-object interaction datasets into 222K multi-object sequences (275 hours, up to 5 objects/sequence), training a single diffusion transformer for human-object interaction that generalizes to unseen objects.

Human-Object InteractionDiffusion TransformerComputer VisionDatasets3D
Category: research
Author: @GerardPonsMoll1
Date: 2026-10-07T00:00:00
Duration: 36.5s
Source: https://x.com/GerardPonsMoll1/status/2107858500621574594

Related Videos

Gen-HumanEgo: Open-Source Egocentric Human Data for Robot Learning
4:14

Gen-HumanEgo: Open-Source Egocentric Human Data for Robot Learning

@GenrobotAI · 775 views · 2026-09-23
Egocentric VisionRobot LearningOpen Source
Odyssey-3: A Foundation World Model for Physical Agents
1:37

Odyssey-3: A Foundation World Model for Physical Agents

@odysseyml · 918 views · 2026-09-15
World ModelsFoundation ModelsPhysical AI
AGIBOT Open-Sources Its World 2026 RL Dataset
0:28

AGIBOT Open-Sources Its World 2026 RL Dataset

@spaceandtech_ · 648 views · 2026-09-03
AgiBotEmbodied AIReinforcement Learning
ACE-Data-0 synchronized multimodal embodied dataset
2:15

ACE-Data-0 synchronized multimodal embodied dataset

@yukangcao · 1.4k views · 2026-07-31
ACEACE-Data-0Embodied AI
Cadena: decompile 3D meshes into editable CAD
0:14

Cadena: decompile 3D meshes into editable CAD

@HuggingApps · 330 views · 2026-08-06
3DCADCadena
AWPlanet 3D Virtual Cinematography Mode
0:15

AWPlanet 3D Virtual Cinematography Mode

@dynamicwangs · 281 views · 2026-07-03
3Dvirtual cinematographyAWPlanet

As an Amazon Associate, we earn from qualifying purchases.