adelevett/docling_pp_layout_demo

1---2title: PP-DocLayoutV3 Empirical Parser3emoji: 📄4colorFrom: blue5colorTo: indigo6sdk: gradio7sdk_version: 6.9.08app_file: app.py9pinned: false10license: mit11---12 13# PDF Layout Detection with PP-DocLayoutV314 15Upload any PDF and get a structured breakdown of every element on the page —16titles, body text, tables, figures, formulas, headers, footers, footnotes, and17more — powered by PaddlePaddle's PP-DocLayoutV3 model via the18[docling-pp-doc-layout](https://github.com/DCC-BS/docling-pp-doc-layout)19plugin.20 21Results are displayed as interactive JSON in the browser and can be downloaded22as a `.json` file with one click.23 24## How to use25 261. Click **Source Document** and upload a PDF.272. Click **Run Layout Detection**.283. Inspect the extracted elements in the JSON panel.294. Click **Download JSON** to save the results.30 31## Output format32 33Each detected region is returned as an object with two fields:34 35```json36{37  "type": "SectionHeaderItem",38  "content": "Introduction"39}40```41 42`type` reflects the docling document-model class. The table below maps the43model's raw labels to the types you will see:44 45| Detected region | Output type |46|---|---|47| `doc_title` | `TitleItem` |48| `paragraph_title` | `SectionHeaderItem` |49| `text`, `content`, `abstract`, `aside_text` | `TextItem` |50| `table` | `TableItem` |51| `image`, `chart`, `seal` | `PictureItem` |52| `formula` | `TextItem` (formula) |53| `footnote`, `vision_footnote` | `TextItem` (footnote) |54| `header` | `TextItem` (page header) |55| `footer` | `TextItem` (page footer) |56| `reference`, `reference_content` | `TextItem` |57| `algorithm` | `TextItem` (code) |58 59## Infrastructure60 61| Component | Detail |62|---|---|63| Hardware | ZeroGPU — NVIDIA H200 (70 GB VRAM, shared) |64| Layout model | [`PaddlePaddle/PP-DocLayoutV3_safetensors`](https://huggingface.co/PaddlePaddle/PP-DocLayoutV3_safetensors) |65| Pipeline | [docling](https://github.com/docling-project/docling) ≥ 2.73 + [docling-pp-doc-layout](https://github.com/DCC-BS/docling-pp-doc-layout) |66| SDK | Gradio 6.9.0, Python 3.10 |67