datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
matrixcity_large_folder_denseDenseFusion-1M
[Paper] https://arxiv.org/abs/2407.08303
[GitHub] https://github.com/baaivision/DenseFusion
Introduction
An image is worth a thousand words". Comprehensive image descriptions are essential for multi-modal perception, while images contains various visual elements of different granularities that are challenging to harness.
We propose Perceptural Fusion to integrate the diverse visual perception experts for capturing visual elements and adopt a MLLM as a centric pivot for… See the full description on the dataset page: https://huggingface.co/datasets/BAAI/DenseFusion-1M.Charge-040_0040-DenseTrainCharge-010_0050-DenseDensePose-COCO
Dataset Card for DensePose-COCO
DensePose-COCO is a large-scale ground-truth dataset with image-to-surface correspondences manually annotated on COCO images.
This is a FiftyOne dataset with 33929 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/DensePose-COCO.dresscode_agnostic_and_densepose
DressCode Agnostic & DensePose Dataset
Agnostic images, corresponding masks, and DensePose images for the DressCode dataset.
Information about the usage can be found at:
https://github.com/jiwoohong93/ita-mdt_code
License
The Dress Code Dataset is proprietary to and © Yoox Net-a-Porter Group S.p.A. and its licensors.It is distributed by the University of Modena and Reggio Emilia and is available for non-commercial academic use under the licence terms provided… See the full description on the dataset page: https://huggingface.co/datasets/jiwoohong93/dresscode_agnostic_and_densepose.Charge-020_0020-DenseCharge-040_0040-DenseTestRethinking-CFG-OPD-Dense2Sparse-Dataset
Dense2Sparse benchmark
Paper · Project Page · Code · Checkpoints
The complete 600-clip evaluation benchmark from our study of classifier-free guidance in
on-policy distillation — 6 motion-difficulty buckets × 100, built on
OpenHumanVid.
Self-contained: download this and you can run the evaluation. No OpenHumanVid download, no
annotation pass, no GPU-days of preprocessing.
rgb/
600 mp4
ground-truth clips, 832×480 / 81 frames
reference_image/
600 png
frame 0 of… See the full description on the dataset page: https://huggingface.co/datasets/Cuttle-fish-my/Rethinking-CFG-OPD-Dense2Sparse-Dataset.DenseUAVdensewalk-publicSa2VA-TrainingThis repository contains the code and data for the paper "Sa2VA: Marrying SAM2 with LLaVA for Dense Grounded Understanding of Images and Videos".
🏠 Project Page📜 arXiv
🧑💻 GitHub
Sa2VA is the first unified model for the dense grounded understanding of both images and videos. Unlike existing multi-modal large language models, which are often limited to specific modalities and tasks, Sa2VA supports a wide range of image and video tasks, including referring segmentation and conversation… See the full description on the dataset page: https://huggingface.co/datasets/Dense-World/Sa2VA-Training.dense_object_detection_FiftyOne
Dense Object Detection Dataset (FiftyOne Format)
Dense Object Detection Dataset for open-world/open-vocabulary object detection models like Grounding DINO, OWL-ViT, and GLIP.
Statistics
Images: 8,001
Annotations: 344,079
Categories: 1,038
Loading with FiftyOne
import fiftyone as fo
import fiftyone.utils.huggingface as fouh
dataset = fouh.load_from_hub("shubh303/dense_object_detection_voxel51")
# Launch the app
session = fo.launch_app(dataset)… See the full description on the dataset page: https://huggingface.co/datasets/shubh303/dense_object_detection_FiftyOne.VLMDense-WebVid-CoVRdense-sc-wdshumanpose_denseposefineweb_synth_dense_ocr_with_groundingfineweb_synth_dense_ocr_colors_with_groundingDenseReward
DenseReward Dataset
🌐 Project page · 📄 Paper (arXiv:2607.13033)
This is the dataset for DenseReward: Dense Reward Learning via Failure Synthesis for Robotic Manipulation. It pairs single robot-manipulation frames (and short chronological frame windows) with a scalar task-progress reward in [0.000, 1.000], used to finetune a vision-language reward model.
Models trained on this data:
densereward/densereward-1frame: single-frame reward model — one RGB frame + task text → scalar… See the full description on the dataset page: https://huggingface.co/datasets/densereward/DenseReward.fineweb_synth_dense_ocr_gradients_with_grounding_Bfineweb_synth_dense_ocr_gradients_with_grounding_Awiki_dense_ocr_v2_textfan48-dense-data
fan48-dense-data
At a state: several action chunks proposed from it, and how each one actually ended.
A branch the search dropped was cut off mid-episode, so it is resumed from its own snapshot
and carried to a finish — the action nobody executed still gets an answer to would this
have worked.
321 searches · 12 tasks · 110,877 nodes, each with its own
state and image.
This repo hosts the data. What it means, how it was produced and how to use it live in
the code that wrote it:… See the full description on the dataset page: https://huggingface.co/datasets/mahgoobi/fan48-dense-data.Dense-Evolution-Ising-Tests
🔬 Quantum Phase Transitions, Variational Gradients, and Error Mitigation
This repository contains a rigorous empirical study, raw datasets, and quantum error mitigation protocols executed on Dense Evolution—a high-performance Statevector quantum simulator. Utilizing 64-bit double precision (complex128) and hardware-accelerated static compilation via the JAX XLA engine, this project maps the non-linear physics of the Transverse Field Ising Model (TFIM) and Tight-Binding… See the full description on the dataset page: https://huggingface.co/datasets/Tatopenn/Dense-Evolution-Ising-Tests.libero10-dense-object-mask-ecot
LIBERO-10 Object Masks + ECoT Reasoning Traces
Object-mask annotations and Embodied Chain-of-Thought (ECoT) reasoning traces for
the 379 demonstrations of the LIBERO-10 (libero_10_image) benchmark.
Generated for the CoT-VLA project. Per-object segmentation masks were produced
with interactive SAM2 point/box prompts and bidirectional video propagation; the CoT
reasoning traces were hand-refined per task and re-timed to each episode's actuator
(gripper + motion) signal.
Source… See the full description on the dataset page: https://huggingface.co/datasets/mwnuk/libero10-dense-object-mask-ecot.thinking-cap-tier-lima-dense
Thinking Cap Tier Curricula — LIMA Hyper-Dense Reasoning Alignment Suite (TCS v4)
[!IMPORTANT]
Dataset Release v1.2 (Sept 2026) — Clean Delimiters & Zero-Padding Architecture:
In v1.2, all 5,500 SFT and 2,000 SimPO records have undergone a complete token purge:
Zero <|pad|> batch residues: 100% eliminated across all records.
Zero reasoning leakage into final answers: Deliberation stays strictly inside <think>...</think>, and answers provide direct conclusions.
Native ChatML… See the full description on the dataset page: https://huggingface.co/datasets/Davd-b01/thinking-cap-tier-lima-dense.multixscience_dense_meanThis is a copy of the Multi-XScience dataset, except the input source documents of its train, validation and test splits have been replaced by a dense retriever. The retrieval pipeline used:
query: The related_work field of each example
corpus: The union of all documents in the train, validation and test splits
retriever: facebook/contriever-msmarco via PyTerrier with default settings
top-k strategy: "max", i.e. the number of documents retrieved, k, is set as the maximum number of documents… See the full description on the dataset page: https://huggingface.co/datasets/allenai/multixscience_dense_mean.details_KoboldAI__fairseq-dense-2.7B
Dataset Card for Evaluation run of KoboldAI/fairseq-dense-2.7B
Dataset Summary
Dataset automatically created during the evaluation run of model KoboldAI/fairseq-dense-2.7B on the Open LLM Leaderboard.
The dataset is composed of 64 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train"… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_KoboldAI__fairseq-dense-2.7B.OpenDetection-15K-Dense-v1.0
OpenDetection-15K-Dense-v1.0
OpenDetection-15K-Dense-v1.0 is an object detection dataset built primarily from general, publicly available images, which make up the majority of the input imagery, together with additional publicly available datasets. The dataset contains high-quality object detection annotations generated using modern automated computer vision pipelines, providing structured object labels, confidence scores, and bounding box coordinates for every detected object.… See the full description on the dataset page: https://huggingface.co/datasets/prithivMLmods/OpenDetection-15K-Dense-v1.0.