Embodied AI Dataset

Domestic Service and Cleaning Datasets

Creator: Changsha Langhui Information Technology Co., Ltd.
Keywords: ,Embodied-intelligence datasets,Robot imitation learning,World-model training,Langhui,AItraining data

10,000+ episodes .. nudity video ..do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do-do

episodes10,000+

Apartments/houses/salvas, etc.6 Household type

Land cleaning/cleaning/cleaning/disposal, etc.7. Domestic work

Face/sensitive information blurPrivacy De-identification

Dataset Overview

Definition：Home service and cleaning data sets are smart data sets provided by Long Salang Langhui Information Technology Ltd. The tasks of cleaning, polishing, sorting, sorting and sorting of waste and assigning items to home settings are collected using bare-hand video and wearing-up cameras, covering multiple households, light and interference with family members, and are suitable for domestic service robots.

Dataset Name	Domestic Service and Cleaning Datasets
Total Data Volume	10,000+ episodes
Data Type	Videos, nudity key points, clean areas mask, natural language commands, voice (optional)
Video Specifications	1920×1080 @ 30 FPS
Delivery Format	LeRobot 3.0 / RLDS / HDF5 / JSON Metadata
License Type	Licensed Data or Custom Projects

Core Data Fields

Each piece of data contains complete structured fields, covering all requirements from basic metadata to professional analysis.

Field Category	Typical Fields
Home scene video	Video, household type, room function, light conditions, family interference markers, camera perspective
The key point of naked hands and limbs.	21 Critical points, arm posture, exposure status, operator object mask, nails/skin area, right-hand distinction
Clean Tasks	Paths to sweep, area wipe, pre- and post-recognition, sorting of waste, type of stain, use of cleaning tools, task successful tagging
Natural Language Command	Task description, step tips, user preferences, abnormal feedback and retry commands, context of multiple rounds of dialogue
Metadata and Compliance	Household code, time frame, authorization range, level of privacy de-identification, family members ' presence

Data Sample Showcase

The following is a de-identified structured JSON sample illustrating the typical data format.

Other Organiser
"HOME-2026-0034512",
"home type": "apartment 2br",
"Room": "Living Room,"
"video": "rgb 0034512.mp4", "deepth 0034512.mkv", "fps":
"hand keypoints": {"left": [...], "right": [...], "contact":[0,0,1,0]},
"tasks"...
"label": "wipe table", "start": 12.0, "end": 35.5, "region": "coffee table", "tool": "cloth"
I don't know.
"objects":
"Obj-112", "category": "cup", "moved": true, "final place": "shelf A2"
I don't know.
"nl instruation":
♪ I'm sorry ♪

AI Application Scenarios

Home ServicesRobot Training

Real home-based data training for general-purpose service robots such as ground cleaning, window cleaning, and cleaning.

I understand naked hands.

Hand key points and exposure data are intended to be a model for the identification and action imitation of human intent.

Long-range mission planning

Multistep cleaning and sorting missions to support long-range task decomposition and planning algorithms.

Human collaboration interactive

The natural language instructions and actions are aligned to train family assistants in following up their instructions.

Categorization of waste

Waste classes and drop action data are trained in garbage classification and recycling robots.

Light and householdization

Multi-household, multi-light data improve the performance of family robots in real homes.

Frequently asked questions

Which family tasks are covered?

Common household chores such as covering the cleaning, mopping, cleaning, sorting, sorting, sorting, dressing, etc.

How can privacy be protected?

All videos were unsensitized and face, sensitive text and personal information were blurred or replaced.

Does it contain voice commands?

The text of the audio and natural language command can be used to synchronize the audio and text to support joint speech-visual-action training.

How to Get a Data Sample?

Please contact the Changsha Langhui Information Technology Co., Ltd. DataAssetsAPI Platform.