Object Detection
Transformers
ONNX
Chinese
English
document-ai
document-layout-analysis
patent
pdf
hiro
patsnap
Instructions to use PatSnap/Hiro-Layout with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use PatSnap/Hiro-Layout with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("object-detection", model="PatSnap/Hiro-Layout")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("PatSnap/Hiro-Layout", dtype="auto") - Notebooks
- Google Colab
- Kaggle
File size: 2,379 Bytes
fc08c11 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 | # Open Source Release Checklist
Use this checklist before publishing Hiro-Layout to Hugging Face or GitHub.
## Repository Metadata
- [ ] Confirm final public model name and repo id, for example `PatSnap/Hiro-Layout`.
- [ ] Confirm model task and Hugging Face `pipeline_tag`.
- [ ] Confirm model architecture, parameter count, input resolution, and output schema.
- [ ] Confirm whether the release includes weights, inference code, configs, examples, and evaluation assets.
- [ ] Confirm whether `layout_model/RT-DETR_25.onnx` is the final public model artifact.
- [ ] Confirm all large binary files are tracked with Git LFS.
## Legal and License
- [ ] Confirm Apache-2.0 is approved for this model and code release.
- [ ] Confirm model weights can be released under the same license or document a separate model license.
- [ ] Confirm training data, evaluation data, and benchmark summaries are cleared for public disclosure.
- [ ] Confirm the Excel benchmark file can be publicly shared.
- [ ] Review `NOTICE` for trademark language.
- [ ] Review `DISCLAIMER.md` for product, legal, and compliance requirements.
## Model Card
- [ ] Replace the minimal ONNXRuntime inspection snippet with the final working inference API.
- [ ] Add installation instructions.
- [ ] Add hardware and runtime requirements.
- [ ] Add preprocessing details for PDF rendering and image normalization.
- [ ] Add output schema, including bounding box format and confidence score semantics.
- [ ] Confirm `labels.json` matches the class-id order used by `layout_model/RT-DETR_25.onnx`.
- [ ] Add example image and example prediction if public samples are available.
- [ ] Confirm benchmark numbers in `README.md`, `README_zh.md`, and `EVALUATION.md`.
## Release Assets
- [ ] Add model weights, config, tokenizer/processor files, and custom code if needed.
- [ ] Add `requirements.txt`, `pyproject.toml`, or environment instructions.
- [ ] Add minimal smoke-test script.
- [ ] Add citation metadata if there is a paper, blog, or technical report.
- [ ] Add a changelog or release notes.
## Final Validation
- [ ] Clone the public repo into a clean environment.
- [ ] Run the documented installation steps.
- [ ] Run the documented inference example.
- [ ] Verify README links render correctly on Hugging Face.
- [ ] Verify the license badge and model metadata render correctly on Hugging Face.
|