AI Training Datasets

diverseactions8hv3 是一套面向 NE01 24-DoF 人形机器人的成对动作数据。每个样本同时提供重定向后的 GMR robot-motion 和重定向前、已规范化到 GMR 坐标系的 SMPL 数据，并通过根目录的 manifest.csv 一一关联。 - trainingsets/ne01sonicmixed455550hzv1：R14 双线训练的冻结输入，4555 对 Robot/SMPL.pkl，50 Hz，含逐文件 SHA-256 和配对 manifest。这里的 SMPL 是训练时配对的 zlib 压缩 joblib.pkl，并非下述 GMR 规范化.npz；读取方法见版本目录说明。 - robotmotion/diverseterrainfree9hv4 与 smpl/diverseterrainfree9hv4：既有 4877 对 GMR 数据。 - robotmotion/diverseactions8hv3 与 smpl/diverseactions8hv3：既有 4376 对 GMR 数据。根目录 manifest.csv、datasetsummary.json 及下文统计描述这一版本。 新训练集在数据集查看器中使用 ne01sonicmixed455550hzv1 配置；原有 diverseactions8hv3 配置保持不变。 相同 sampleid 的.pkl 和.npz 表示同一个动作。manifest.csv 同时给出两侧路径、文件大小、SHA-256、是否存在以及 pairstatus。 本版本按 contactfree…

Dataset · Image to image

Noob2EDIT

AnimeNexa LaxCore

HF 仓库：AnimeNexaLaxCore/Noob2EDIT，公开、人工审核开放。仓库内 data/ 下是分片、旁路表和 manifest.json，demo/ 是加载器适配用的演示包。 来源：ModelScope niangao233/NOOB2EDITGAMECG15M（hitomi gamecg 画廊，按相似度聚类成组）。 本目录是清洗、配对、VLM 标注后的训练用导出，格式为 WebDataset 分片，一个样本 = 一个差分组。 demo/ 下是给数据加载器适配用的小规模演示包，格式与正式分片完全一致。 - family：cg 游戏 CG 差分；sprite 立绘（原图带透明通道，已合成白底）。两套分片分开，训练时按比例混合。 - split：train / val，按画廊 id 的 sha1 哈希切分（val 占 2%），同一游戏的组不会跨集。 - 分片目标大小 1 GiB（正式），demo/ 为 300 MB。分片内的组顺序按画廊哈希打乱，各年份、各游戏天然混合。 - 分片文件名与 index 单调递增，manifest 中记录每个分片的字节数、sha256、样本数、边数和旁路表 sha256。 一个组在 tar 中是连续的若干成员，key 为 {galleryid}g{组号}： - CG 分支保留原始 AVIF 字节，未重编码。 - 立绘和任何带真实透明像素的图，在写入时合成到纯白背景并重编码为 AVIF q90（images[].reencoded = true）。训练数据中不存在 alpha 通道。 - 组内所有图尺寸一致（width /…

Dataset · Other

copart-dataset

Rebrowser.net

prettyname: "Copart Salvage Vehicle Auction Dataset" - open-data - data-science - web-scraping - scraper - crawler - data-collection - copart - salvage - auction - vehicles - automotive - insurance - total-loss - copart-com sizecategories: taskcategories: - other - configname: auction-listings datafiles: Salvage vehicle auction listings with damage assessments, condition grades, title status, and yard locations from Copart's nationwide network. This dataset is a preview sample of the Copart dataset published by Rebrowser. If you're doing academic research, you may be eligible for free access to a much larger slice — see Free Datasets for Research. This dataset contains 1 entity, each in its…

ORCA-3DCT — precomputed 3D CT visual tokens & Data Encoder outputs for the whole of CT-RATE, released so that work on 3D CT token compression, pooling, retrieval and probing does not have to start by spending several hundred GPU-hours re-encoding the corpus. ctrate/uncompressed/ — the raw encoder grids, plus the pooled encoders' native output. Bring your own compressor. ctrate/compressed/ — ORCA and grid-average token

Dataset

users-and-organizations

Yağız Çalık

This dataset contains usernames and organization names in a TXT file. This dataset is automatically updated using GitHub Actions.

Warning: This is a pre-release version. Labels in this dataset are subject to change without notice. This is a marine mammal vocalisation dataset primarily focused on the southern resident killer whale habitat. The portion of this data comes from Ocean Networks Canada, and is shared under a CC-BY License. Alternatively, the data can be downloaded from Ocean Networks Canada Directly: Marine mammal detections are conducted by an amature labeller with a Jaccard score of ~0.7 compared to an exper bioacoustician. Species and ecotype-level labels are not compared. If you use this, please cite: As well as the specific DOIs for the subset of ONC data you use located in DORI.csv

Dataset

klein-layout-plans

Yuvraj Sharma

Training and evaluation data for two small layout planners (Qwen3.5-0.8B and Qwen3.5-2B) distilled from the FLUX.3 Image Klein prompt upsampler (black-forest-labs/flux-3-image-klein-pup) in its t2ibbox mode. Status: stub — data lands here as the gated pipeline progresses. This file is rewritten with the real card when the dataset is complete. - 8k–10k image requests (taxonomy of multi-object, counting, text/poster, infographic, portrait, product and single-subject scenes; ~15% non-English; ~30% with a TARGET CANVAS line), with structured specs for ~40% of them. - Teacher outputs (single-line caption + JSON box table) from the pup's t2ibbox mode, with finish reason and token counts.…

Per-frame metric scene depth (Depth-Anything-V2 Metric Indoor Large, 20 m ceiling) for the episodes inpainted in VR-VLA/VR-egodex-inpaint-cache-full. Depth is the second input Eq. 4 compositing needs beside the inpainted plate; together they let scripts/egodex/buildtask.py retarget an episode onto all 16 embodiments. Archive names are IDENTICAL to the inpaint repo, one tar per /: Episode numbers match the source dataset's episode%06d.mp4. is part of the key: most task names exist in both train/ and test/ with overlapping episode numbers. Archives are uncompressed tar (PNG payload). train/addremovelidv2.1.tar (308 GB, 1,158 episodes) exceeds the Hub's 50 GB file limit and is stored as…

Shopify stores detected from public Certificate Transparency logs, usually within an hour of their custom domain's first HTTPS certificate (median 0.4 h, measured on 4,168 stores on 2026-10-05), often while the store is still behind its password page. Source: public Certificate Transparency logs, public DNS and RDAP, and each store's own public storefront metadata. The radar samples those logs, so it does not catch every new Shopify store. 981 of them (16 %) pre-launch at their last status check. Top countries: US (2142), FR (429), GB (363), IN (317), ES (288), PK (260). Top niches: General store (986), Fashion & Apparel (633), Home & Garden (421), Beauty & Skincare (341), Jewelry & Watches…

Dataset · Video text to text

finevideo-256

Tristen Jackson

Derived from HuggingFaceFV/finevideo (https://huggingface.co/datasets/HuggingFaceFV/finevideo). Videos are Creative Commons; original licenses apply, including attribution where required. Provenance for every clip is in its.json (provenance, sourceid). must update to the latest usable version (use reshardvideo.py purge). Hosting or sharing requires including the FineVideo Terms of Use and requiring users to agree. If this is hosted publicly it must be gated, with these terms shown and accepted by every user. Modifications: cut into short clips, resized, re-encoded (H.264), audio removed. FineVideo dataset is a collection of over 43.000 YouTube videos. We ask that you read and acknowledge…

Private, append-only research data collected by hourlycryptoquant. - futurewindowl21s: contracts, snapshots - polymarketl21s: contracts, featuresnapshots - polymarketl25s: contracts, featuresnapshots - polymarketrtds: priceticks, structuralsnapshots - externalvenuel21s: venueseconds Rows are exported from SQLite as immutable Parquet shards. The repository is updated approximately every four hours. Paths are partitioned by source, table, and UTC date. rowid is the source-database row identifier used for incremental checkpointing. This dataset intentionally excludes environment files, credentials, wallet and account information, live orders, live executions, paper portfolios, logs, and…