datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
InternData-A1
InternData-A1
InternData-A1 is a hybrid synthetic-real manipulation dataset containing over 630k trajectories and 7,433 hours across 4 embodiments, 18 skills, 70 tasks, and 227 scenes, covering rigid, articulated, deformable, and fluid-object manipulation.
Your browser does not support the video tag.
Your browser does not support the video tag.… See the full description on the dataset page: https://huggingface.co/datasets/InternRobotics/InternData-A1.A12d12s12TIIF-Bench-DataWe release the images generated by the proprietary models evaluated in “🔍TIIF-Bench: How Does Your T2I Model Follow Your Instructions?”.
Produced under carefully crafted, high-quality prompts, these images form a valuable asset that can benefit the open-source community in a variety of applications🔥.
MICo-BenchInter-Edit-Test
Inter-Edit-Test
Official test benchmark release for the CVPR 2026 paper:
Inter-Edit: First Benchmark for Interactive Instruction-Based Image Editing
This repository hosts the public release of Inter-Edit-Test, a human-annotated benchmark for the Interactive Instruction-based Image Editing (I^3E) task.
Each sample contains:
a source image,
a coarse user-style interaction mask,
a concise editing instruction,
and a ground-truth edited image.
To simplify large-scale distribution on… See the full description on the dataset page: https://huggingface.co/datasets/a1557811266/Inter-Edit-Test.bedroom-tidying-robot-manipulation-training-set-next-pack-93682a70-a19b80d2
Bedroom furniture perception — attic & flat-ceiling rooms
Training datapack to improve a robot perception model on bedroom scenes furnished with a double bed, two nightstands, a dressing table and a wardrobe, across two configurations: attic/sloped-ceiling bedrooms (skylight windows) and flat-ceiling bedrooms. 30 renders at 1024x1024 with RGB plus albedo, metric depth and world-space OpenGL normals, per-frame annotations enabled and each environment's authored lighting. Supports… See the full description on the dataset page: https://huggingface.co/datasets/physicl-community/bedroom-tidying-robot-manipulation-training-set-next-pack-93682a70-a19b80d2.RAD_DataSet
RAD Dataset (Remove/Add Dataset)
A large-scale, fully synthetic dataset for image editing tasks, containing 514,510 high-quality annotated image tuples.
Overview
The RAD dataset is generated through a rigorous three-stage pipeline:
Textual Prompt Generation: Structured guidance prompts generated using Qwen3-32B LLM
Image Pair Synthesis: Diptych image pairs synthesized using FLUX.1 diffusion transformer, fine-tuned evolutionarily via LoRA
Quality Filtering &… See the full description on the dataset page: https://huggingface.co/datasets/a1557811266/RAD_DataSet.pack-54o-of-my-project-53d7si-a11b689a
Outdoor Urban Structure Detection Eval
Evaluation dataset of 10 renders across 10 outdoor and urban environments (streets, alleyways, a street cafe, a pocket park, a street library) at 1024x1024, with albedo and metric depth passes and per-frame annotations, to measure a computer vision model's detection of urban structures such as buildings, roads and sidewalks.
This dataset mirrors public data-pack render outputs from Physicl.
Each row represents one render view. The image… See the full description on the dataset page: https://huggingface.co/datasets/physicl-community/pack-54o-of-my-project-53d7si-a11b689a.fall_detectSynCPR
SynCPR Dataset
Overview
The SynCPR dataset is a large-scale, fully synthetic dataset designed specifically for the composed person retrieval task. Built using our automated construction pipeline, SynCPR offers unmatched diversity, quality, and realism for person-centric image retrieval research.
For more details, see https://github.com/Delong-liu-bupt/Composed_Person_Retrieval
Construction Pipeline
The dataset is constructed in three main stages:… See the full description on the dataset page: https://huggingface.co/datasets/a1557811266/SynCPR.2016-6-A1-dataset-prevLAM_Cases_For_ECCVcorpus-A19052024IML_Datasetsadd_local
Local & Add 编辑数据集
本数据集包含图像编辑任务的训练数据，编辑类型为 Local（局部编辑）和 Add（添加元素）。
数据集统计
总样本数: 455,477 条
Local 类型: 286,659 条
Add 类型: 168,818 条
文件结构
add_local/
├── local_add_only.json # 数据标注文件（路径为相对路径）
├── images_part_0000.zip # 图像压缩包 1
├── images_part_0001.zip # 图像压缩包 2
├── ... # 更多压缩包
├── extract_all.py # 批量解压脚本
└── README.md # 本说明文件
快速开始
1. 解压数据
运行解压脚本，将所有压缩包解压到当前目录：
python extract_all.py… See the full description on the dataset page: https://huggingface.co/datasets/a1557811266/add_local.twitter-QQ202202-2026.03.29-2038140372753481983-a1LAqYhksUN6wcEl-part1osworld_tasks_filesmaniskill2-replayed-trajectoriesimmuc8j47g
im
timm-resnset18-a1-in1k-val-top_5-chunkedStable_Diffusion_WebUI_a1111app2016-A1_Jun1app2017-2-A1_Jul6_plotML-A1ZebraMini2016-6-A1-unnormalized-newtwitter-tianxin5572-2025.01.31-1885361193751326967-QDBQl-u7HJv0T-a1-part12016-6-A1-unnormalized2016-6-A1-dataset-newgui360-balanced-a11y