dataset_name
string | last_updated
timestamp[s] | source
string | record_count
int64 | fields
dict | notes
string |
|---|---|---|---|---|---|
VizWiz Visual Question Answering - Validation Annotations
| 2026-01-21T00:00:00
|
VizWiz Challenge (vizwiz.org) - Danna Gurari et al., CVPR 2018
| 4,319
|
{
"image": "Image filename (e.g., 'VizWiz_val_00000000.jpg')",
"question": "Natural language question asked by the visually impaired user about the image",
"answers": "JSON array of 10 crowd-sourced answers, each with 'answer' text and 'answer_confidence' (yes/maybe/no)",
"answer_type": "Category of the answer (e.g., 'unanswerable', 'other', 'yes/no', 'number')",
"answerable": "Binary flag: 1 if the question is answerable from the image, 0 if not"
}
|
Validation split of the VizWiz VQA dataset containing 4,319 image-question pairs. Images were taken by blind or visually impaired users using smartphones. Each question has 10 crowd-sourced answers with confidence ratings. The 'unanswerable' category captures cases where the image quality is too poor or the question cannot be answered from the visual content. Used for accessibility research and assistive technology development.
|
Accessibility Atlas
42 datasets covering disability demographics, employment gaps, web accessibility compliance, assistive technology patents, special education, healthcare, housing discrimination, transportation, government benefits, and more -- across 15 categories.
Built for researchers, journalists, policy analysts, and anyone building tools for disabled communities.
What's Inside
US Disability Demographics (Census Bureau)
8 datasets from the American Community Survey covering county, state, and national disability rates.
- County-level: 3,100+ counties (ACS 5-Year 2022)
- State-level: 50 states + DC (ACS 5-Year 2021)
- National trends: 13-year trend (11.9% to 13.6%)
- Demographics: by race, age, sex, and 6 disability types
Disability Employment
4 datasets on the employment gap, discrimination charges, and workplace accommodations.
- BLS employment/unemployment by disability status
- FRED 16-year time series (gap narrowing from 40pp to 37pp)
- 679,637 ADA discrimination charges (1992-2024)
- JAN accommodation cost data (58.7% cost $0)
Special Education (IDEA)
6 datasets from the National Center for Education Statistics on 7.5M students served under IDEA Part B.
- 13 disability categories
- 51 states
- 1976-2023 historical data
- Demographics, environments, state-level breakdowns
Web Accessibility
5 datasets on WCAG compliance, lawsuits, screen reader usage, and Section 508.
- WebAIM Million: 94.8% of top 1M pages fail WCAG
- Screen reader survey: 1,539 respondents
- ADA digital lawsuits: 814 (2017) to ~4,000 (2024)
- Section 508: 245 federal entities, only 23% conform
- HTTP Archive a11y trends 2019-2024
International Disability
8 datasets covering 194+ countries.
- WHO: disability prevalence, DALYs/YLDs, healthy life expectancy
- OECD: 34 countries, prevalence/employment/spending
- Eurostat: 36 EU countries
- World Bank: 10 socioeconomic indicators
- UN CRPD: 199 countries ratification status
Government Benefits
- SSA: SSDI/SSI annual data (7.3M + 4.9M beneficiaries)
- VA: disability compensation by state (5.7M+ veterans)
Healthcare & Mental Health
- CMS: Medicaid disability enrollment (10.2M enrollees)
- SAMHSA: mental health prevalence 2008-2023
Housing, Transportation, Education
- HUD: fair housing complaints (disability is #1 basis, ~55%)
- NTD: paratransit ridership/costs ($40-50/trip)
- CRDC: Section 504, restraint/seclusion, discipline
Assistive Technology Patents
- WIPO: patent data 1985-2026 (10 → 519 patents)
Visual Accessibility
- VizWiz: 4,319 VQA pairs from blind users (CC BY 4.0)
Key Findings
44.1 million Americans (13.4%) have a disability (Census 2022). Rate rose from 11.9% (2010) to 13.6% (2023). Ambulatory (6.7%) and cognitive (5.7%) are most common.
The employment gap is massive. 24.5% labor force participation (disabled) vs ~67% (non-disabled). OECD average gap: 34 percentage points.
Workplace accommodations are cheap. 58.7% cost nothing. Median: $300. Yet 679,637 ADA charges filed since 1992.
The web is still broken. 94.8% of top 1M websites fail WCAG. Average: 51 errors/page.
Disability is #1 basis for fair housing complaints (~55% of all HUD complaints).
Special education is growing. 15.2% of public school students now served under IDEA. Autism grew 8.5x since 2000.
Globally, 1.3 billion people (16%) live with significant disability (WHO). 190 countries signed UN CRPD.
Notebooks
10 Jupyter notebooks in notebooks/ that load the data, run analysis, and produce charts:
census_disability_demographics.ipynb: Maps, trends, demographic breakdownsdisability_employment.ipynb: Employment gap over time, discrimination chargesweb_accessibility.ipynb: WCAG failure trends, screen reader market shareinternational_disability.ipynb: Cross-country comparisonsspecial_education_idea.ipynb: Enrollment trends, discipline disparitiesgovernment_benefits.ipynb: Benefit trends, state-level processinghealthcare_mental_health.ipynb: Enrollment, treatment gapshousing_transportation_education.ipynb: Fair housing, transit costsassistive_technology.ipynb: Patent growth by categorysign_language_vision_aac.ipynb: Visual question answering
Plus accessibility_atlas_demo.ipynb in root as a quick-start sampler.
Data Sources & Licenses
Most data is from US government (public domain) and international organizations (open licenses). Individual datasets carry their own licenses:
| Source | License |
|---|---|
| US Census Bureau | Public Domain |
| Bureau of Labor Statistics | Public Domain |
| US Dept. of Education | Public Domain |
| SSA, VA, CMS, SAMHSA | Public Domain |
| HUD, NTD, CRDC | Public Domain |
| WHO Global Health Observatory | CC BY-NC-SA 3.0 IGO |
| OECD, Eurostat, World Bank | Open licenses (see sources) |
| WebAIM | Fair use (structured summaries) |
| VizWiz | CC BY 4.0 |
Full source table with URLs in main README.
File Format
Data files: JSON, CSV, Excel (XLSX) Notebooks: Jupyter (.ipynb)
About
Part of the Data Trove collection. I built this because disability data is scattered across dozens of federal agencies, international organizations, and advocacy groups. Having it all in one place -- with consistent formatting and analysis notebooks -- makes it actually usable.
Luke Steuber • lukesteuber.com • @lukesteuber.com
Citation
@misc{steuber2026accessibility,
title={Accessibility Atlas: Disability Demographics and Web Accessibility Dataset Collection},
author={Steuber, Luke},
year={2026},
url={https://huggingface.co/datasets/lukeslp/accessibility-atlas}
}
License
MIT License for the collection. Individual datasets have their own licenses (see table above). Public domain data can be used freely. WHO data is CC BY-NC-SA 3.0 IGO. VizWiz is CC BY 4.0.
- Downloads last month
- 47