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
| # 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. | |