2024 · Preprint
Unlocking the conversion of Web Screenshots into HTML Code with the WebSight Dataset
WebSight is a synthetic dataset of 2 million HTML documents paired with full-page screenshots for training screenshot-to-HTML models. Mistral-7B-Instruct generates website concepts, Deepseek-Coder-33B-Instruct writes each page in HTML with Tailwind CSS, images come from keyword-based stock-photo URLs, and Playwright renders the screenshots after filtering. This v0.2 release replaces the 823K-pair v0.1, which used traditional CSS and colored rectangles instead of images. The authors fine-tune an 8B vision-language model, Sightseer, on the data with DoRA and show qualitative conversions, including one from a hand-drawn sketch.
Scope and limitations
The websites are synthetic, LLM-generated single pages, not crawled real sites, so training on them does not establish reconstruction of real websites, their source or their behavior. Evaluation is qualitative only: no benchmark scores are reported, checkpoints were selected by manual inspection because validation loss did not track output quality, and the authors note failures on complex layouts, long text and designs unlike the training data, as well as Tailwind syntax errors. v0.2 pages load Tailwind CSS from a CDN link and images from source.unsplash.com URLs, so re-rendering depends on those external services. The base model was unreleased at the time, and the report is not peer-reviewed.
Why it fits this collection
Adds the synthetic screenshot-to-HTML training dataset that later website reconstruction work, including Design2Code and Web2Code, builds on.
Publication and review details
- Venue
- arXiv
- Status
- Preprint
- Publication / posting date
- March 14, 2024
- Date source
- arXiv v1 submission date; no later version, journal reference or proceedings publication was found.
- First arXiv submission
- March 14, 2024
- arXiv
- 2403.09029
- Reviewed arXiv version
- v1
Reviewed arXiv v1, the only version, a technical report with no comment, journal reference or DOI; a Crossref search found no peer-reviewed version. The project link pins the Hugging Face dataset card at revision b11f817 (last modified 2024-03-26). The card declares CC BY 4.0, asks users to follow the licenses of the source content and to disclose use when releasing trained models, documents v0.1 in detail and still marks the v0.2 split size and creation section “TO DO”. The paper does not link code or the Sightseer weights, so no code link is recorded. The dataset was not downloaded, and no code or experiments were run.
Sources reviewed
Source review only; experiments were not rerun for this collection. Cite the original authors using the linked paper or BibTeX.