Skip to content

Dataset · iciclab

Dual-camera video and speech for robot interaction

iciclab/Robot_physics_interaction_with_Human_linguistic

300 sessions featuring first- and third-person video, speech audio, and text action labels.

What's inside

The dataset consists of 300 recorded human-robot interaction sessions. Each folder contains synchronized first- and third-person videos (MP4), a speech audio track (WAV), a text transcript of the read script, and a JSON action label file.

What you can do with it

  • Training multimodal systems for understanding voice commands and human actions in physical spaces
  • Developing automatic speech recognition (ASR) and visual grounding models for service robotics
Downloads this month
20.8K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("iciclab/Robot_physics_interaction_with_Human_linguistic")