Rapid Layout Document Layout Analysis

Rapid Layout is a document layout analysis tool that identifies the category and position of layout elements such as titles, paragraphs, tables, and images in document-type images (e.g., paper screenshots, research reports).

Supported Scenarios:

• Tables
• Chinese text
• English text
• Academic papers
• Research reports
• General layouts

The project integrates various open-source models including the PP series, YOLOv8 series, and DocLayout-YOLO. Because different scenarios have significant layout variations, it is recommended to fine-tune with a custom dataset if the built-in models do not meet your specific business requirements.

To use the OpenVINO engine, first install the required packages: pip install rapid-layout onnxruntime openvino. You can then initialize RapidLayout directly by setting engine_type=EngineType.OPENVINO. Configuration for OpenVINO devices and threads can be managed via the openvino section in the engine configuration.

pip install rapid-layout onnxruntime openvino

from rapid_layout import EngineType, ModelType, RapidLayout

layout_engine = RapidLayout(
    model_type=ModelType.PP_LAYOUT_CDLA,
    engine_type=EngineType.OPENVINO,
)
img_path = "https://raw.githubusercontent.com/RapidAI/RapidLayout/718b60e927ab893c2fad67c98f753b2105a6f421/tests/test_files/layout.jpg"
results = layout_engine(img_path)
print(results)
results.vis("layout_res.png")

All RapidLayout models are hosted on the ModelScope platform. You can download them from the following repository:

RapidLayout Models on ModelScope

Install the rapid-layout package along with onnxruntime using pip to enable layout analysis capabilities.

pip install rapid-layout onnxruntime

Tests are located in the tests/ directory. Use pytest from the repository root to verify your changes.

• Run all tests: pytest tests/ -v
• Run a specific test file: pytest tests/test_main.py -v
• Check test coverage: pytest tests/ -v --cov=rapid_layout (requires pytest-cov installed)

# Run all tests
pytest tests/ -v

# Run specific file
pytest tests/test_main.py -v

# Check coverage
pytest tests/ -v --cov=rapid_layout

The following environment is required for converting PP-DocLayout models to ONNX:

• OS: Ubuntu
• Python: 3.10.14
• Dependencies:
  • paddle2onnx==2.1.0
  • paddlepaddle==3.3.0
  • onnx==1.17.0
  • onnxruntime==1.23.2

Install the core library and the default onnxruntime engine using pip:

pip install rapid-layout onnxruntime

If you want to use the openvino engine, you must also install OpenVINO:

pip install openvino

Use the paddle2onnx command to convert the model directory and files to an .onnx file. Ensure you provide the correct paths to the model directory, the JSON file, and the parameter file.

paddle2onnx  --model_dir=models/PP-DocLayoutV2  --model_filename inference.json --params_filename inference.pdiparams  --save_file=./models/PP-DocLayoutV2/inference.onnx  --enable_onnx_checker=True

When adding new features or fixing bugs, you must include corresponding unit tests.

Guidelines:

• Place test files in tests/ with the naming convention test_*.py.
• Use pytest for writing test cases.
• Place test resources (like images) in tests/test_files/.
• Ensure tests are stable and do not depend on undocumented external services (use mocks where necessary).

Example test structure:

# tests/test_xxx.py
import pytest
from pathlib import Path

cur_dir = Path(__file__).resolve().parent
root_dir = cur_dir.parent
test_dir = cur_dir / "test_files"

def get_engine():
    from rapid_layout import RapidLayout
    return RapidLayout()

def test_your_new_feature():
    engine = get_engine()
    img_path = test_dir / "layout.jpg"
    result = engine(img_path)
    assert result is not None
    # Add more assertions...