Repository navigation
CellHunter v0.1.0-alpha
Pre-releaseCellHunter 0.1.0-alpha
Developer preview. This is the first public alpha of CellHunter, a Windows
desktop application for automated cell counting in microscopy images. The
desktop app and the Cellpose analysis pipeline work end to end, but the project
is still an MVP: interfaces, report columns, and packaging may change in later
releases.
Highlights
- Batch analysis of a flat image folder with Cellpose 3.0.7 (
nuclei). - Automatic GPU/CPU selection with a safe CPU fallback.
- Live progress bar and streamed analyzer logs.
- Per-image results table: file, nucleus count, dimensions, processing time.
- Excel report written next to the analyzed images.
- Self-contained Windows build: no .NET installation required.
What is in the archive
CellHunter.Desktop.exe self-contained WPF application (win-x64)
analyzer/ Python analyzer package (main.py)
requirements.txt pinned Python dependencies
README.md full project documentation
LICENSE MIT license
The Python analyzer is not bundled. Cellpose and PyTorch are large
dependencies, so the application looks for a local Python environment at
runtime.
Requirements
| Component | Requirement |
|---|---|
| OS | Windows 10/11 (64-bit) |
| Python | 3.10–3.12 (64-bit); 3.12 recommended |
| .NET | Not required for this build |
| GPU | Optional NVIDIA GPU with CUDA |
| Disk | ~5 GB for the Python environment and model weights |
Quick start
-
Unpack the archive into a writable folder, for example
C:\CellHunter. -
Create the Python environment in the unpacked folder:
cd C:\CellHunter py -3.12 -m venv venv venv\Scripts\python -m pip install --upgrade pip venv\Scripts\pip install -r requirements.txt
-
Start
CellHunter.Desktop.exe. -
Select a folder with images, pick the
cellposemodel, and click Запуск
(Start). The user interface is currently in Russian.
For CUDA acceleration, install a CUDA-enabled PyTorch build from
pytorch.org before step 2, then run
the remaining requirements as shown above.
The first analysis downloads the pretrained Cellpose nuclei weights and
therefore requires an internet connection.
Known limitations
- Only top-level files in the selected folder are analyzed; subfolders are not
traversed. - The GUI has no device selector: it requests GPU execution and falls back to
CPU automatically. - The YOLO option is experimental and has no bundled cell-detection weights.
- The Excel button in the UI is a placeholder; reports are generated
automatically after each run. - Images in non-ASCII paths (for example Cyrillic) cannot be read by OpenCV on
Windows; keep input paths ASCII-only. - The executable is not code-signed, so Windows SmartScreen may show a warning.
- The build was verified by compiling and packaging; a full functional pass on
representative microscopy data has not been performed for this alpha.
Verification
Download SHA256SUMS.txt next to the archive and verify the checksum:
Get-FileHash .\CellHunter-0.1.0-alpha-win-x64.zip -Algorithm SHA256Feedback
Please report issues, including the log output and image modality, at
https://github.com/bat801/CellHunter/issues.
License
MIT. See LICENSE.