CVF Papers Scraper

Extract research papers, authors, PDFs, BibTeX citations, and abstracts from the Computer Vision Foundation Open Access repository — covering CVPR, ICCV, and WACV.

📖 What does it do?

The CVF Papers Scraper extracts structured metadata from openaccess.thecvf.com, the official open-access repository of the Computer Vision Foundation. It covers papers from the three major CVF-sponsored conferences:

• CVPR (Conference on Computer Vision and Pattern Recognition) — annually since 2013, the largest CV conference
• ICCV (International Conference on Computer Vision) — biennially (odd years) since 2013
• WACV (Winter Conference on Applications of Computer Vision) — annually since 2020

What you get for each paper:

• Full paper title
• Authors list (as both an array and a formatted string)
• Conference name and year
• Direct PDF download URL
• Supplemental materials URL (when available)
• arXiv preprint URL (when available)
• Complete BibTeX citation
• Page numbers from the proceedings
• Full abstract text (optional — requires one extra request per paper)

All data is extracted directly from the static HTML — no JavaScript rendering required, no login, no authentication.

---

👥 Who is it for?

🎓 Computer vision researchers and PhD students

Building a literature review on a CV topic? Tracking what got accepted at CVPR 2024? Use this actor to bulk-download paper metadata instead of clicking through thousands of entries manually. Filter by conference and year to get exactly the batch you need.

📊 Scientometricians and bibliometrics analysts

Studying publication trends in computer vision, measuring author collaboration networks, tracking how research topics evolve across CVPR/ICCV/WACV — start with structured, machine-readable paper data covering 10+ years of the field's most prestigious venues.

🤖 AI and ML practitioners building datasets

Creating fine-tuning corpora from CV abstracts, building citation graphs, or constructing benchmarks from paper lists — this actor delivers clean JSON at scale with no manual effort.

🏢 Research teams and enterprise R&D labs

Monitoring what competitors or academic collaborators are publishing, building internal research intelligence dashboards, or feeding paper data into RAG (retrieval-augmented generation) systems for literature QA.

📚 Academic librarians and information professionals

Maintaining curated databases of CV research, populating institutional repositories, or building subject guides — all with properly formatted BibTeX citations and direct PDF links.

🛠️ Developers building research tools

Creating paper recommendation engines, topic clustering tools, author disambiguation systems, or academic search interfaces — this actor gives you the raw structured data to build on.

---

🚀 Why use it?

• Fast and cheap — openaccess.thecvf.com is fully server-side rendered with no anti-bot measures. One HTTP request per conference fetches all paper listings. No browser needed, no proxy required.
• Complete coverage — CVPR 2013–2025 (~100,000 papers total), ICCV 2013–2025 (~40,000 papers), WACV 2020–2026 (~15,000 papers)
• BibTeX included — Every paper's complete citation is available on the listing page, ready to paste into your reference manager
• arXiv cross-linking — Where available, the arXiv preprint URL is extracted so you can access unrestricted versions
• Abstract fetching — Enable includeAbstract to get full abstract text from each paper's detail page
• Structured output — Clean JSON with typed fields (year as integer, authors as array), ready for downstream processing

---

📊 Data fields extracted

Field	Type	Description
title	string	Full paper title
authors	string[]	Author names as an array
authorsString	string	Authors joined as a single comma-separated string
conference	string	Conference code (CVPR, ICCV, or WACV)
year	number	Conference year (e.g., 2024)
pages	string | null	Page range in proceedings (e.g., "4864-4873")
paperUrl	string	URL to the paper detail page on CVF open access
pdfUrl	string | null	Direct URL to the PDF file
suppUrl	string | null	URL to supplemental materials (PDF or ZIP)
arxivUrl	string | null	arXiv preprint URL (when listed)
bibtex	string | null	Complete BibTeX citation string
abstract	string | null	Full abstract text (only when includeAbstract: true)

---

💰 Pricing

This scraper uses Pay-Per-Event (PPE) pricing — you pay only for papers actually extracted, not for compute time.

What you pay for	FREE	BRONZE	SILVER	GOLD	PLATINUM	DIAMOND
Run started (one-time)	$0.005	$0.005	$0.005	$0.005	$0.005	$0.005
Each paper extracted	$0.00115	$0.00100	$0.00078	$0.00060	$0.00040	$0.00028

Cost examples (at BRONZE $0.001/paper):

• 100 papers (quick test or small workshop): ~$0.10
• 1,000 papers (single conference track): ~$1.01
• 5,000 papers (CVPR main track): ~$5.01
• 15,000 papers (full CVPR 2024): ~$15.01

With abstract fetching (includeAbstract: true): each paper requires one additional HTTP request, but the PPE price stays the same — only time and network overhead increase slightly.

Free plan: Apify's free tier includes $5 of monthly credit, enough for ~4,300 papers per month at BRONZE pricing with no credit card required.

---

🛠️ How to use it

Step 1: Choose conferences and years

Configure which conferences and years to scrape. You can mix and match:

{
    "conferences": ["CVPR", "ICCV"],
    "years": [2023, 2024]
}

This would scrape CVPR 2023, CVPR 2024, ICCV 2023, and ICCV 2024 (4 batches total).

Step 2: Set a result limit

Use maxResults to control cost and run time. Start with a small number (e.g., 50) to test:

{
    "conferences": ["WACV"],
    "years": [2024],
    "maxResults": 50
}

Set to a large value (e.g., 999999) for unlimited results.

Step 3: Optionally fetch abstracts

Enable includeAbstract to retrieve the full abstract from each paper's detail page:

{
    "conferences": ["CVPR"],
    "years": [2024],
    "includeAbstract": true,
    "maxResults": 100
}

Note: Abstract fetching is ~10× slower because it requires one HTTP request per paper.

Step 4: Run and download

Click Start and wait for completion. Download results as JSON, CSV, or Excel from the Dataset tab.

---

📥 Input parameters

conferences — Which conferences to scrape

Array of conference codes. Options: CVPR, ICCV, WACV. Default: ["CVPR"].

{ "conferences": ["CVPR", "ICCV", "WACV"] }

Pass "all" as a string shorthand to scrape all three conferences for the selected years.

Available years per conference:

• CVPR: 2013–2025 (annual)
• ICCV: 2013, 2015, 2017, 2019, 2021, 2023, 2025 (biennial, odd years only)
• WACV: 2020–2026 (annual)

If you specify a year that isn't available for a conference (e.g., ICCV 2024), it will be skipped with a log message.

years — Which years to scrape

Array of integer years. Default: [2024].

{ "years": [2022, 2023, 2024] }

maxResults — Limit output size

Maximum number of papers to extract across all selected conferences and years. Default: 500.

{ "maxResults": 1000 }

Set to a large value (e.g., 999999) for unlimited. Papers are extracted in listing order (top to bottom on the CVF page).

includeAbstract — Fetch abstract text

When true, the actor fetches each paper's detail page to extract the full abstract. Default: false.

{ "includeAbstract": true }

This makes the run slower (one extra HTTP request per paper) but enables full-text search, text analysis, and semantic similarity use cases.

maxRequestRetries — Retry count

Number of retry attempts for failed HTTP requests. Default: 3. Range: 1–10.

{ "maxRequestRetries": 5 }

---

📤 Output examples

Paper without abstract

{
    "title": "DPMesh: Exploiting Diffusion Prior for Occluded Human Mesh Recovery",
    "authors": ["Yixuan Zhu", "Ao Li", "Yansong Tang", "Wenliang Zhao", "Jie Zhou", "Jiwen Lu"],
    "authorsString": "Yixuan Zhu, Ao Li, Yansong Tang, Wenliang Zhao, Jie Zhou, Jiwen Lu",
    "conference": "CVPR",
    "year": 2024,
    "pages": "1101-1110",
    "paperUrl": "https://openaccess.thecvf.com/content/CVPR2024/html/Zhu_DPMesh_Exploiting_Diffusion_Prior_for_Occluded_Human_Mesh_Recovery_CVPR_2024_paper.html",
    "pdfUrl": "https://openaccess.thecvf.com/content/CVPR2024/papers/Zhu_DPMesh_Exploiting_Diffusion_Prior_for_Occluded_Human_Mesh_Recovery_CVPR_2024_paper.pdf",
    "suppUrl": "https://openaccess.thecvf.com/content/CVPR2024/supplemental/Zhu_DPMesh_Exploiting_Diffusion_CVPR_2024_supplemental.zip",
    "arxivUrl": "http://arxiv.org/abs/2404.01424",
    "bibtex": "@InProceedings{Zhu_2024_CVPR,\n    author    = {Zhu, Yixuan and Li, Ao and ...},\n    title     = {DPMesh: Exploiting Diffusion Prior for Occluded Human Mesh Recovery},\n    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},\n    month     = {June},\n    year      = {2024},\n    pages     = {1101-1110}\n}",
    "abstract": null
}

Paper with abstract

{
    "title": "Seeing the World through Your Eyes",
    "authors": ["Hadi Alzayer", "Kevin Zhang", "Brandon Feng", "Christopher A. Metzler", "Jia-Bin Huang"],
    "authorsString": "Hadi Alzayer, Kevin Zhang, Brandon Feng, Christopher A. Metzler, Jia-Bin Huang",
    "conference": "CVPR",
    "year": 2024,
    "pages": "4864-4873",
    "paperUrl": "https://openaccess.thecvf.com/content/CVPR2024/html/Alzayer_Seeing_the_World_through_Your_Eyes_CVPR_2024_paper.html",
    "pdfUrl": "https://openaccess.thecvf.com/content/CVPR2024/papers/Alzayer_Seeing_the_World_through_Your_Eyes_CVPR_2024_paper.pdf",
    "suppUrl": null,
    "arxivUrl": "http://arxiv.org/abs/2306.09348",
    "bibtex": "@InProceedings{Alzayer_2024_CVPR,...}",
    "abstract": "The reflections in the eyes contain information about the environment around the person, including the appearance of the illumination and objects in the room..."
}

---

💡 Tips and tricks

• Fast survey mode: Skip includeAbstract (default false) to scrape a full conference in under 60 seconds — the listing page has all metadata except abstract text.
• Abstract-enriched NLP: Enable includeAbstract: true when you need full text for topic modeling, semantic search, or LLM analysis. Expect ~10× longer run time for large batches.
• Sampling a conference: Set maxResults: 50 to get a random sample — great for testing your downstream pipeline before committing to a full 2,000+ paper run.
• Multi-year longitudinal studies: Use years: [2019, 2020, 2021, 2022, 2023, 2024] with conferences: ["CVPR"] to get six years of papers in one run.
• Use arXiv links for full text: Many papers have arxivUrl populated, giving you the preprint even without IEEE Xplore access.
• Import BibTeX directly: The bibtex field is a complete, valid BibTeX entry — paste directly into your .bib file or any reference manager.
• Deduplication key: Use paperUrl as your unique identifier when combining results across multiple runs.
• ICCV is biennial: ICCV only runs in odd years (2021, 2023, 2025). Specifying an even year returns 0 results; the actor skips it with a log message.

---

🔗 Integrations

📊 Google Sheets research tracker

1. Run the actor with conferences: ["CVPR"], years: [2024], and maxResults: 500
2. Open Google Sheets → Extensions → Apify (requires the Apify Sheets add-on)
3. Import the dataset and map columns: title, authorsString, conference, year, arxivUrl, pdfUrl
4. Result: a live spreadsheet of all selected papers, sortable and filterable

📚 Zotero / Mendeley BibTeX bulk import

1. Run with conferences: ["CVPR", "ICCV"], years: [2024], includeAbstract: false
2. Download the JSON dataset
3. Extract BibTeX fields into a .bib file using Python: python3 -c "import json; [print(p['bibtex']) for p in json.load(open('papers.json')) if p['bibtex']]" > cvf2024.bib
4. Import cvf2024.bib into Zotero or Mendeley — all papers import with full structured metadata

🤖 LLM trend analysis with Claude or ChatGPT

1. Run with includeAbstract: true on a focused batch (e.g., 100 papers from CVPR 2024)
2. Feed the JSON to an LLM with the prompt: "Analyze these CVPR 2024 paper abstracts and identify the 10 dominant research themes with examples"
3. Automate this pipeline using Apify's Claude or OpenAI integrations

🔍 Semantic search with Pinecone / Weaviate

1. Scrape all ICCV 2023 papers with includeAbstract: true
2. Embed abstracts using the OpenAI Embeddings API (or run the Apify OpenAI Embeddings actor)
3. Push vectors to Pinecone with paper metadata — instantly queryable: "find papers similar to NeRF"

📈 Publication trend dashboard in Tableau / Power BI

1. Schedule a monthly Apify run scraping the latest conference year
2. Connect the Apify dataset via the REST API to your BI tool
3. Build trend charts: papers per year, top authors, arXiv adoption rate, keywords in titles

---

🔌 API usage

You can trigger this actor programmatically using the Apify API.

Python

import apify_client

client = apify_client.ApifyClient("YOUR_APIFY_TOKEN")

run = client.actor("automation-lab/cvf-papers-scraper").call(
    run_input={
        "conferences": ["CVPR"],
        "years": [2024],
        "maxResults": 100,
        "includeAbstract": False
    }
)

for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["title"], "-", item["authorsString"])

cURL

curl -X POST \
  "https://api.apify.com/v2/acts/automation-lab~cvf-papers-scraper/runs" \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "conferences": ["CVPR"],
    "years": [2024],
    "maxResults": 100
  }'

Node.js

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });

const run = await client.actor('automation-lab/cvf-papers-scraper').call({
    conferences: ['CVPR'],
    years: [2024],
    maxResults: 100,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(`Extracted ${items.length} papers`);

---

🤖 MCP (Model Context Protocol)

Use this actor as an MCP tool inside Claude, Cursor, VS Code, or any MCP-compatible AI assistant to fetch CVF papers on demand from natural language prompts.

Claude Code (terminal)

$claude mcp add --transport http apify "https://mcp.apify.com?tools=automation-lab/cvf-papers-scraper"

Claude Desktop / Cursor / VS Code

Add to your claude_desktop_config.json (or equivalent MCP config):

{
    "mcpServers": {
        "apify": {
            "type": "http",
            "url": "https://mcp.apify.com?tools=automation-lab/cvf-papers-scraper",
            "headers": {
                "Authorization": "Bearer YOUR_APIFY_TOKEN"
            }
        }
    }
}

Example prompts for your AI assistant

Once connected, you can ask:

• "Get the 50 most recent CVPR 2024 papers about 3D reconstruction"
• "Scrape all ICCV 2023 papers and summarize the dominant research themes"
• "Fetch WACV 2024 papers with abstracts — I want to find work on medical imaging"
• "Pull all CVPR papers from 2022 to 2024 and export their BibTeX citations for my literature review"

---

⚖️ Legality and ethical use

openaccess.thecvf.com provides papers under open access as part of the Computer Vision Foundation's mission to advance research. The data extracted by this actor is:

• Publicly available — all papers are freely accessible without login or authentication
• Non-commercial research data — paper metadata (titles, authors, abstracts) is factual bibliographic information, not copyrighted content
• Explicitly intended for distribution — the CVF open access repository exists specifically to make this research accessible

Usage guidance:

• Use the maxRequestRetries setting to avoid hammering the server with excessive retries
• CVF's open access terms permit non-commercial research use of paper metadata
• The full PDF content of papers is copyrighted by authors/IEEE — downloading PDFs at scale may require separate permissions
• This actor extracts metadata only by default; always check the CVF terms of service and your organization's policies before commercial use

We do not encourage scraping beyond your legitimate research needs.

---

❓ FAQ

Q: Does this require a proxy? No. openaccess.thecvf.com has no anti-bot measures. The actor uses plain HTTP requests — no residential proxy needed.

Q: How many papers are available total? Approximately 100,000 CVPR papers (2013–2025), 40,000 ICCV papers, and 15,000 WACV papers — around 155,000 total across all supported conferences.

Q: Can I scrape ECCV (European Conference on Computer Vision)? ECCV is not hosted on openaccess.thecvf.com — it's organized separately. This actor covers only CVF-sponsored conferences (CVPR, ICCV, WACV).

Q: Why does ICCV only have odd years? ICCV is a biennial conference held in odd-numbered years. CVPR is annual. WACV became annual starting in 2020.

Q: What happens if I specify a year that isn't available? The actor logs a warning and skips that conference/year combination. Other valid combinations continue normally.

Q: Is abstract fetching significantly more expensive? The PPE price per paper is the same whether or not you fetch abstracts. But with includeAbstract: true, the run takes longer because of extra HTTP requests — roughly 10× longer for large batches.

Q: Can I get papers from a specific research topic or keyword? CVF open access doesn't have a filtering API — papers are listed alphabetically. Use maxResults to cap output, then filter locally by title or abstract content.

Q: The actor seems to have duplicate papers — why? This shouldn't happen in normal operation. Each paper appears once per listing page. If you scrape the same conference/year in multiple runs, you'll get duplicates across datasets — deduplicate by paperUrl.

---

🔗 Related actors

• ACL Anthology Scraper — Extract NLP and computational linguistics papers from aclanthology.org (ACL, EMNLP, NAACL, EACL, COLING, and 50+ workshops)
• ArXiv Scraper — Search and download paper metadata from arXiv.org across all subject areas — ideal for tracking preprints that correspond to CVF papers
• Semantic Scholar Scraper — Retrieve citation counts, author profiles, and paper abstracts from Semantic Scholar's AI-powered research database

---

🐛 Issues and feedback

Found a bug or have a feature request? Open an issue on GitHub or contact us through the Apify support channel. We actively maintain this actor and welcome pull requests.

Common issues:

• Missing papers for a specific year: Verify the year is in the supported range for that conference
• Slow run with abstracts: Abstract mode fetches one extra page per paper — expected behavior
• arXiv URL missing: Not all CVF papers have arXiv preprints; those fields will be null