datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
4DThinker-Training-Data
4DThinker Training Data
This repository contains the training data for 4DThinker, a framework that enables VLMs to "think with 4D" through dynamic latent mental imagery, built upon SpatialVID and DSR_Suite-Data.
Data Structure
data/
├── dift_data.jsonl # DIFT training data (~38K samples)
├── 4drl_data_filtered.jsonl # 4DRL training data (~37K samples)
└── processed_data/ # Video frames & mask overlays
├── <video_id>/
│ ├── frames/… See the full description on the dataset page: https://huggingface.co/datasets/jankin123/4DThinker-Training-Data.4D-Lung
4D-Lung (segmentation subset)
Longitudinal 4D (respiratory-gated, phase-resolved) fan-beam CT of 20
locally-advanced non-small-cell lung cancer (NSCLC) patients, with expert
manual RTSTRUCT contours, from Data from 4D Lung Imaging of NSCLC Patients
(4D-Lung) on The Cancer Imaging Archive (Hugo et al., VCU).
This is the segmentable subset of the full collection — read carefully.
The full TCIA 4D-Lung collection is 183 GB and contains both 4D fan-beam CT
(4D-FBCT, "4DCT") and 4D… See the full description on the dataset page: https://huggingface.co/datasets/MedOtter/4D-Lung.4DNeX-10M
4DNeX-10M Dataset
📄 Paper | 🚀 Project Page | 💻 GitHub
Introduction
4DNeX-10M is a large-scale hybrid dataset introduced in the paper "4DNeX: Feed-Forward 4D Generative Modeling Made Easy".
The dataset aggregates monocular videos from diverse sources, including both static and dynamic scenes, accompanied by high-quality pseudo 4D annotations generated using state-of-the-art 3D and 4D reconstruction methods. The dataset enables joint modeling of RGB appearance and… See the full description on the dataset page: https://huggingface.co/datasets/3DTopia/4DNeX-10M.MVOIK-4D
HAT4D: Human-Assisted Training for 4D Dynamic Scene Understanding
MVOIK-4D: Multi-View Object Interaction Knowledge for 4D Physical Reasoning
If you find this dataset useful, please consider citing our paper and following the project page for updates.
💡 Description
MVOIK-4D is a curated release of real-world object-interaction sequences for 4D scene understanding, reconstruction, and evaluation. It contains RGB input frames, multi-view evaluation frames, and memory-mask… See the full description on the dataset page: https://huggingface.co/datasets/Lijiaxin0111/MVOIK-4D.iPhone360-4dgs360
iPhone360 Dataset - 4dgs360 preprocessed version
iPhone360 is a benchmark dataset for 360° reconstruction of dynamic objects from monocular video, introduced in the paper:
4DGS360: 360° Gaussian Reconstruction of Dynamic Objects from a Single Video
Jae Won Jang, Yeonjin Chang, Wonsik Shin, Juhwan Cho, Nojun Kwak
Project Page · arXiv
Dataset Description
iPhone360 features real-world dynamic scenes captured with an iPhone, where test cameras are positioned at… See the full description on the dataset page: https://huggingface.co/datasets/mipal/iPhone360-4dgs360.Lattice_4D_Dataset
Lattice_4D_Dataset
Multi-camera volumetric captures of people doing everyday tasks (ball
handling, shirt folding). Four synchronised RGB-D cameras record each take.
Each take has the reconstructed 3-D scene of every frame and the fitted body
and hands. It also has calibration, camera poses, per-frame action labels,
reviewed language and rendered orbit videos. A USD skeleton and a URDF rig let
a robotics consumer load the body.
The action orbit of dataset_balls_p1_2 with the… See the full description on the dataset page: https://huggingface.co/datasets/latticecx/Lattice_4D_Dataset.captain_cook_4drv-4d54f178b0
Internal video store for RuLips-1k
This repository is the working storage behind
levossadtchi/RuLips-1k:
the actual 224×224 clips collected in August–September 2026, kept for our own
training. It is not a curated dataset: folder names are collection machines,
some clips are duplicates, and only RuLips-1k/manifest.jsonl says which files are
part of the release (store.path / store.member / store.batch per clip).
Layout: lips-*/batchNNNNN/<video>/spanNNN_tK_cut.mp4 and bigN/... —… See the full description on the dataset page: https://huggingface.co/datasets/levossadtchi/rv-4d54f178b0.polytopes-4d
Four-Dimensional Reflexive Lattice Polytopes
This dataset contains all four-dimensional reflexive lattice polytopes. The data was
compiled by Maximilian Kreuzer and Harald Skarke in
arXiv:hep-th/0002240. More information is
available at the Calabi-Yau data website.
Please cite the paper when referencing this dataset:
@article{Kreuzer:2000xy,
author = "Kreuzer, Maximilian and Skarke, Harald",
title = "{Complete classification of reflexive polyhedra in four-dimensions}"… See the full description on the dataset page: https://huggingface.co/datasets/calabi-yau-data/polytopes-4d.4dgen-datasetcite: arxiv.org/abs/2507.01099
cv4cdd_4d
Content
This repository stores the contents of the data/ directory from the following GitLab repository:https://gitlab.uni-mannheim.de/processanalytics/cv4cdd
The data is organized as follows:
input_cdlgContains training, validation, and test datasets used to train the computer vision models.
input_cdriftContains external datasets used to evaluate the trained models.
model_training_loggingContains model checkpoints for all training runs. This includes both relevant checkpoints… See the full description on the dataset page: https://huggingface.co/datasets/pm-science/cv4cdd_4d.4DReasoner_v4_test4d3ac703c0b98d419a7ce729a504eb897401435fGBI-16-4D
GBI-16-4D Dataset
GBI-16-4D is a dataset which is part of the AstroCompress project. It contains data assembled from the Sloan Digital SkySurvey (SDSS). Each FITS file contains a series of 800x800 pixel uint16 observations of the same portion of the Stripe82 field, taken in 5 bandpass filters (u, g, r, i, z) over time. The filenames give the
starting run, field, camcol of the observations, the number of filtered images per timestep, and the number of timesteps. For example:… See the full description on the dataset page: https://huggingface.co/datasets/AstroCompress/GBI-16-4D.4D-LRM-Stuffsft_env_b9057b9c-a10c-4d1d-a360-4c79e5201dcac5_eng_nfp_fine_4dumpsZJU-4DRadarCamABot-World-Explorer-4D
ABot World Explorer 4D
ABot World Explorer 4D is a depth-enabled sample of the action-conditioned
video data infrastructure described in
ABot-World-0. Its source manifest references
20 episodes and 181,561 EXR depth objects; the release preserves their bytes.
Dataset facts
Item
Value
Episodes
20
Base source objects
120
EXR depth objects
181,561
Total source objects
181,681
Semantic splits
None
Depth representation
Absolute metric… See the full description on the dataset page: https://huggingface.co/datasets/acvlab/ABot-World-Explorer-4D.VSL-4D-237wpublic-4dstem
ECLIPSE-Lab public 4D-STEM hub
Published, experimental 4D-STEM datasets from other groups, re-hosted in one
uniform, lossless-compressed HDF5 format (e4d) with complete, verified
calibration metadata, for benchmarking reconstruction, compression and
denoising methods across a wide dose range (ptychography and nanobeam
diffraction). Every dataset remains the work of its original authors:
please cite the original publication (see Citations below).
What every file… See the full description on the dataset page: https://huggingface.co/datasets/philipp-pelz/public-4dstem.4DReasoner_v3_part2XianKa-Cat-4D-Data
XianKa-Cat-4D-Data
两位抖音博主的视频数据集 + SAM 2.1 语义分割结果，按博主（抖音号/UID）分别打包。
本数据集不包含 4D 重建点云——请使用 4RC 对视频自行运行 4D 重建。
内容
文件/目录
说明
28680959431.tar
博主"显卡从"（抖音号 28680959431）：1342 个视频 + metadata.json
28680959431_seg.tar
该博主的 SAM 2.1 分割结果
96187538465.tar
博主"史官"（抖音号 96187538465）：569 个视频 + metadata.json
96187538465_seg.tar
该博主的 SAM 2.1 分割结果
examples/
每位博主 10 个样例（视频 + 对应分割），平铺可直接在线浏览
manifest.json
打包清单与统计
视频文件名为 序号_作品ID.mp4；metadata.json 中 aweme_id 与作品 ID… See the full description on the dataset page: https://huggingface.co/datasets/XianKa-Zhong/XianKa-Cat-4D-Data.NV-Raw2insights-MRI-4DFlow-DevMLLM-4D-Datasets
MLLM-4D-Datasets
Project Page | Paper | GitHub
MLLM-4D-Datasets is the datasets introduced in the paper "MLLM-4D: Towards Visual-based Spatial-Temporal Intelligence".
It contains large-scale 4D instructional data such as MLLM4D-2M and MLLM4D-R1-30K, and is designed for advancing the visual-based spatial-temporal intelligence of MLLMs.
Dataset Usage
The MLLM-4D-Datasets can be downloaded using the following command, as indicated in the official GitHub repository:… See the full description on the dataset page: https://huggingface.co/datasets/flow666/MLLM-4D-Datasets.GBI-16-4D
GBI-16-4D Dataset
GBI-16-4D is a dataset which is part of the AstroCompress project. It contains data assembled from the Sloan Digital SkySurvey (SDSS). Each FITS file contains a series of 800x800 pixel uint16 observations of the same portion of the Stripe82 field, taken in 5 bandpass filters (u, g, r, i, z) over time. The filenames give the
starting run, field, camcol of the observations, the number of filtered images per timestep, and the number of timesteps. For example:… See the full description on the dataset page: https://huggingface.co/datasets/AnonAstroData/GBI-16-4D.mbgfnet-closed-shell-4d5d-gw-dataset
Closed-shell 4d/5d transition-metal complexes: PBE0 + G0W0 quasiparticle dataset
2,210 closed-shell mononuclear 4d and 5d transition-metal complexes (17 metals: Y, Zr, Nb, Mo, Ru,
Rh, Pd, Ag, Cd from the 4d row; Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg from the 5d row), each with a DFT
(PBE0/cc-pVDZ, relativistic small-core pseudopotential on the metal) and a one-shot G0W0@PBE0
quasiparticle-energy calculation. Built to extend MBGF-Net
(Venturella, Li, Hillenbrand, Zhu… See the full description on the dataset page: https://huggingface.co/datasets/primateria/mbgfnet-closed-shell-4d5d-gw-dataset.Music-POSTPROCESS-4d73ca94MVISTA-4D
MVISTA-4D
Multi-view 4D robot-manipulation dataset spanning three sources — RobotWin (simulation),
RLBench (simulation), and a real-robot collection. Each episode provides synchronized
multi-view RGB + depth video together with robot joint/action and camera parameters.
The raw data is stored as four tar shards (raw_data.00–03.tar.part, about 122 GB total),
rather than as individually browsable files. Use the provided helper to download, verify, and
extract the data.… See the full description on the dataset page: https://huggingface.co/datasets/ethenj/MVISTA-4D.c5_eng_nfp_nemo_4dumps