datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
scene_splat_7k
Updates
The HyperSimGS and ReplicaGS data have been updated with improved quality.
The ScanNetGS and ScanNetPPGS-V2 data have been updated with improved quality.
The ARKitScenes 3DGS data have been updated with improved quality.
The Matterport3D 3DGS data have been updated with improved quality. The new GS files were optimized from each complete scene rather than from a region. The preprocessed npy version is available at Matterport3D 3DGS. Note that the test_eval folder… See the full description on the dataset page: https://huggingface.co/datasets/GaussianWorld/scene_splat_7k.real2sim-sample-usdz-scenes
Niantic Spatial Real2Sim Sample USDZ Scenes
Sample real-world USDZ scenes and particle-field variants from Niantic Spatial for robotics simulation and physical AI workflows.
Why this exists
The real world is the best simulation.
This repository contains sample scene assets from Niantic Spatial to help developers evaluate real-to-sim robotics and physical AI workflows in NVIDIA Isaac Sim and Isaac Lab.
What's included
This repository includes two… See the full description on the dataset page: https://huggingface.co/datasets/NianticSpatial/real2sim-sample-usdz-scenes.humancentric-scenes-ai
HumanCentric-Scenes-AI
A multimodal benchmark of 296 AI-generated human-centric scenes across four domains:
CCTV / surveillance imagery (Set 2, 85 images). Midjourney-generated stills that mimic low-resolution security-camera footage — parking lots, building interiors, outdoor public spaces — designed to test whether detection cues survive heavy compression and low-light noise.
Occupation × gender portraits (Set 3, 128 images). A balanced 64-occupation × 2-gender paired design… See the full description on the dataset page: https://huggingface.co/datasets/Nima0Kamali/humancentric-scenes-ai.humancentric-scenes-ai
HumanCentric-Scenes-AI
A multimodal benchmark of 296 AI-generated human-centric scenes across four domains:
CCTV / surveillance imagery (Set 2, 85 images). Midjourney-generated stills that mimic low-resolution security-camera footage — parking lots, building interiors, outdoor public spaces — designed to test whether detection cues survive heavy compression and low-light noise.
Occupation × gender portraits (Set 3, 128 images). A balanced 64-occupation × 2-gender paired design… See the full description on the dataset page: https://huggingface.co/datasets/rjmaftv33/humancentric-scenes-ai.scene_splat_49k
Update
We have updated the ARKitScenes 3DGS for improved quality.
We have updated the Matterport3D 3DGS data part for improved quality, which are optimized from the full scenes. Note test_eval folder saves 3DGS corresponds to the test split of the region_segmentations version.
SceneSplat-49K
We present SceneSplat-49K, a large-scale 3D Gaussian Splatting dataset comprising approximately 49K raw scenes and 45K curated 3DGS scenes, aggregated from multiple… See the full description on the dataset page: https://huggingface.co/datasets/GaussianWorld/scene_splat_49k.viet-robot-music-scenes
viet-robot-music-scenes
A small synthetic dataset of music scenes and preferences for a home robot.
Columns:
id: row id
scene_id: identifier of the scene
description_vi: description in Vietnamese
description_en: description in English
time_of_day: morning, afternoon, evening, night
mood: calm, focused, energetic, sleep
volume_level: 1–10 suggested volume
Manually authored for demo purposes.
License
MIT
Breaking_Bad_Scenes_LLMurban-scenes-v2
Urban Scenes – Preprocessing Utilities
This repository contains helper scripts used during preprocessing and validation
of the Urban Scenes v2 dataset.
Contents
preprocess.py – feature normalization utilities
stats.ipynb – exploratory statistics
label_map.json – class label mapping
Notes
Raw data is not included due to licensing constraints.
viet-robot-weather-scenes
viet-robot-weather-scenes
A tiny synthetic dataset of weather scenes and suggested indoor robot modes.
Columns:
id: row id
weather: sunny, cloudy, raining, storm, hot, cold
time_of_day: morning, afternoon, evening, night
outside_temp_c: approximate outdoor temperature
outside_humidity_pct: approximate outdoor humidity
suggested_mode: normal, quiet, safe_check_windows, extra_dry, stay_docked
note_en: short English note
For demo and educational purposes only.
License… See the full description on the dataset page: https://huggingface.co/datasets/hoangs/viet-robot-weather-scenes.viet-robot-light-scenes-v2
viet-robot-light-scenes
A small synthetic dataset of preferred lighting scenes for a smart home.
Columns:
id: row id
scene_id: identifier of the lighting scene
room: room name
time_of_day: morning, afternoon, evening, night
activity: short description
brightness_pct: 0–100
color_temp_k: approximate color temperature in Kelvin
note_en: short English note
For demo and educational purposes only.
License
MIT
viet-robot-light-scenes
viet-robot-light-scenes
A small synthetic dataset of lighting scenes for a smart home robot.
Columns:
id: row id
scene_id: short identifier of the lighting scene
description_vi: description in Vietnamese
description_en: description in English
room: room where the scene is typically used
brightness: brightness level (0-100)
color_temp: color temperature in Kelvin (e.g., 2700, 4000)
Manually authored for demos.
License
MIT