Napari Hydra Plugin - GUI-based nested object detection
Project description
Napari-Hydra
A napari plugin for single-shot nested instance segmentation of biomedical objects using HydraStarDist — a branched deep learning architecture built on StarDist.
Napari-Hydra enables GUI-based detection, segmentation, and quantification of spatially correlated, star-convex objects (e.g. viral plaques nested within tissue culture wells) directly from digital photographs — no microscopy required.
How to Cite Us
If you use this plugin or the underlying methods in your research, please cite the following:
De, T., Thangamani, S., Urbański, A., & Yakimovich, A. (2025). A digital photography dataset for Vaccinia Virus plaque quantification using Deep Learning. Scientific Data, 12, 719. https://doi.org/10.1038/s41597-025-05030-8
De, T., Urbanski, A., & Yakimovich, A. (2025). Single-shot Star-convex Polygon-based Instance Segmentation for Spatially-correlated Biomedical Objects. arXiv preprint arXiv:2504.12078. https://doi.org/10.48550/arXiv.2504.12078
Overview
Traditional virological plaque assays rely on manual counting — a process that is laborious, subjective, and error-prone. Napari-Hydra addresses this by providing an end-to-end deep learning pipeline accessible through napari's graphical interface.
What is HydraStarDist?
HydraStarDist (HSD) extends the StarDist architecture with a joint encoder and branched decoders, enabling simultaneous detection of two categories of nested objects in a single forward pass. This is in contrast to conventional approaches that require independent models for each object type (e.g. one for wells, one for plaques).
The shared encoder implicitly captures spatial correlations between nested objects (e.g. plaques can only appear within wells), resulting in more meaningful representations and improved joint detection accuracy.
Key Features
- Single-shot prediction — detect and segment both wells and plaques simultaneously
- Interactive thresholding — tune probability and NMS thresholds per object class
- Per-well plaque counting — automatic quantification with per-well breakdown
- Model fine-tuning — adapt pre-trained models to your own data directly from the GUI
- Stack support — process time-lapse image stacks with frame-by-frame results
- Export — save prediction summaries including plaque counts and morphometrics
Installation
pip install napari-hydra
Or install in development mode:
git clone https://github.com/plaque2/napari-hydra.git
cd napari-hydra
pip install -e ".[test]"
Usage
Launch napari with the plugin:
napari -w napari-hydra
- Load an image — drag and drop or use
File > Opento load a plaque assay photograph. - Run Prediction — select the image layer and model, then click Run Prediction. Wells and plaques are detected simultaneously and displayed as label overlays.
- Count — click Count to compute per-well plaque counts.
- Tune — fine-tune the model on your own annotated data using the Tune Model button. Fine-tuned models are saved as new timestamped copies, preserving the original.
- Export — save a prediction summary with plaque counts and areas.
License
BSD-3-Clause
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file napari_hydra-0.0.4.tar.gz.
File metadata
- Download URL: napari_hydra-0.0.4.tar.gz
- Upload date:
- Size: 68.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.9.25
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
aeac45f36802d2899411291b3ca184c4e62e5f890dca458c2216e18c512e56fd
|
|
| MD5 |
b16a509fe24ec74488068394c5db4ecf
|
|
| BLAKE2b-256 |
4ce5f42b000000cae32975e6b4b506b363a04759df749ca003b691861f0813bc
|
File details
Details for the file napari_hydra-0.0.4-py3-none-any.whl.
File metadata
- Download URL: napari_hydra-0.0.4-py3-none-any.whl
- Upload date:
- Size: 71.4 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.9.25
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
a77dfab299442f3856f3bd5b6f85a2174b2675add180ecca74c27a40343ffee0
|
|
| MD5 |
b8c73121da6309915ffed83fe060c79a
|
|
| BLAKE2b-256 |
dd3dcd4995f9c02b5dcf3ae1fec1572c63e003163d97f07199fb695416107bd0
|