AI Training Datasets

Dataset · Image classification

COCO_val14_160x120_small

Omnivision

This dataset is a customized subset of the Visual Wake Words (VWW) dataset, specifically optimized for downstream TinyML applications. Visual Wake Words represents a common micro-controller use case: determining whether a person is present in the frame (a "wake word" for vision applications). VWW dataset is entirely derived from the Microsoft COCO 2014/2017 dataset splits, filtered and re-labeled into a binary classification task. This dataset features a split-licensing structure inheriting from its upstream source, MS COCO: 1. Annotations: Any new or modified annotations, labels, and bounding boxes created for this project are released under the Creative Commons Attribution 4.0…

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference A web-based interface for preparing audio datasets to fine-tune OpenAI's Whisper model. This tool helps in recording, managing, and organizing voice recordings with their corresponding transcriptions, with support for cloud storage and authentication. - ⌨ Keyboard shortcuts for efficiency 1. Create a transcript CSV file with your content: 2. Start the Flask application: 3. Access the interface: 1. Authentication 2. Session Setup - Click "Start Session" 3. Recording - Use on-screen controls or keyboard shortcuts: - R: Start recording / Stop recording - Space: Play recording - Enter: Save…

Dataset · Other

tennis-3d-synthetic

Neil

This repository currently contains a bounded storage-format qualification, not a completed training dataset. No production release is complete or ready for use. The qualification lives on branch qualification/s02-format-v2, under qualification/s02-format-v2/. Its manifest.incomplete.json explicitly records state: incomplete. The qualification converts six existing simulated rallies into one numerical Parquet shard without resimulation. It contains 12 views and 2,503 view frames: train 4 rallies / 8 views / 1,218 frames; validation 1 / 2 / 1,080; test 1 / 2 / 205. These are historical development fixtures, not a new production split. Do not use them as a production evaluation benchmark. The…

Dataset

mesh-1t-prod-mix

Chutes

Canonical archive of the parallax 1T-token production curriculum, phase-split, EXACTLY as prod runs it. NOT a text dataset: artifacts are flat llama3 uint32 token bins (EOS 128001 between docs), val bins (doc-level 1/100), plus per-pool mixmanifest.json / poolconfig.json provenance and SHA256SUMS. Files >45GB are stored as.part-NNN chunks (cat name.part- > name to reassemble; per-part sha256 in CHUNKS.sha256). (4a) -> 5e6-8e6 or YaRN (4b). Phase-4 replay packed with document-boundary masking ON. Realized per-source shares, exact token counts, floors, sampling seeds, decontamination (13-gram, stride 4) and every assembly decision/deviation: see MANIFEST.json and phase34mixmanifest.json.…

This repository contains three environment variants of Each variant contains the same 4,578 tasks in five Parquet shards, with the original 14-column schema and task order. Only dockerimage is replaced with the corresponding public image containing the selected agent. Images are hosted in loongsage/r2e-gym on Docker Hub. Choose one configuration for a run; combining configurations repeats the same tasks. The task description is problemstatement. The original prompt column is the meta-prompt used to generate the task, and expectedoutputjson contains the original grading information. All original columns are retained, including commit contents, execution results, modified files and entity…

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference A web-based interface for preparing audio datasets to fine-tune OpenAI's Whisper model. This tool helps in recording, managing, and organizing voice recordings with their corresponding transcriptions, with support for cloud storage and authentication. - ⌨ Keyboard shortcuts for efficiency 1. Create a transcript CSV file with your content: 2. Start the Flask application: 3. Access the interface: 1. Authentication 2. Session Setup - Click "Start Session" 3. Recording - Use on-screen controls or keyboard shortcuts: - R: Start recording / Stop recording - Space: Play recording - Enter: Save…

Object-detection dataset of 22 household appliances, built to train the detector of FixHome, a SEP490 capstone project at FPT University: a marketplace that connects customers with home-repair technicians and gives a preliminary, AI-suggested diagnosis from a photo. The detector (YOLO11s) names the appliance and draws its box; the box crops the photo for a vision-language model further down the pipeline. The dataset leans towards appliances as they look in Vietnamese homes. Public datasets mostly show Western hardware: US and EU wall sockets where Vietnam uses two-pin universal sockets, built-in ranges where Vietnam uses countertop gas burners, front-loading washers where Vietnam is largely…

A durability-focused research corpus of AI-generated creative workflow media — node-graph session snapshots and their associated rendered assets (thumbnails, stills, and encoded previews) collected from public community galleries for generative content platforms. Media corpora for generative workflows are commonly stored on ephemeral CDNs with high link-rot. This corpus packages a snapshot of such assets in a sharded, integrity-checked, encrypted archival format suitable for durability/replication research and offline analysis of generative content characteristics (encoding formats, size distributions, thumbnail vs full-asset trade-offs). - shards/.data — encrypted (AES-256-CTR) tar…