annovert provides the functionalities for converting between different annotation formats used in digital pathology.
from pathlib import Path
from annovert import Annovert, ConverterConfig, Format
config = ConverterConfig(output_directory=Path("output"))
converter = Annovert(config)
converter.convert_file(
input_path="annotation.xml",
output_path="output/annotation.geojson", # optional if config set
target_format=Format.GEOJSON_QUPATH_MRXS,
image_path="slide.mrxs"
)
converter.convert_file(
input_path="annotation.xml",
output_path="output/mask.tiff", # or .jpg, .tif, .png
target_format=Format.BINARY_MASK,
image_path="slide.mrxs",
downsample_factor=8.0
)
converter.batch_process(
input_dir="annotations/",
output_dir="output/", # optional if config set
target_format=Format.BINARY_MASK,
image_dir="images/",
downsample_factor=4.0
)Format |
Value | Extension | Description |
|---|---|---|---|
Format.ASAP |
"asap" |
.xml |
ASAP XML annotations |
Format.GEOJSON |
"geojson" |
.geojson |
Standard GeoJSON |
Format.GEOJSON_QUPATH_MRXS |
"geojson_qupath_mrxs" |
.geojson |
GeoJSON with MRXS offset/mpp metadata¹ |
Format.HISTECH_3D |
"3dhistech" |
.xml |
3D Histech / CaseViewer XML (source only) |
Format.BINARY_MASK |
"binary_mask" |
.png .jpg .tif |
Binary segmentation mask¹ |
Format.HAMAMATSU_NDPA |
"hamamatsu_ndpa" |
.ndpa |
Hamamatsu NanoZoomer NDPA XML |
¹ Requires image_path or image_dir for slide metadata
Plain strings are accepted everywhere for backward compatibility ("geojson_qupath_mrxs" is equivalent to Format.GEOJSON_QUPATH_MRXS).
Either direct from Github:
pip install git+https://github.com/computationalpathologygroup/annovert.gitor from source
git clone git@github.com:computationalpathologygroup/annovert.git
cd annovert
pip install -e .