MIRA - Microscopy Image Recognition and Analysis

MIRA is a professional PyQt5 application designed for real-time spore detection and quantification using deep learning (YOLO) and microscopy imaging.

---

📖 Documentation

Full documentation is available at mira-microscopy.readthedocs.io

The documentation includes:

• Detailed installation guides for all platforms.
• Comprehensive usage tutorials.
• Hardware configuration and camera settings.
• Advanced data analysis and export workflows.

---

Key Features

• 🔬 Real-time Analysis: Live video feed with YOLO-based object detection.
• 🧪 Standardized Chambers: Built-in support for Malassez and KOVA slides.
• 📊 Automated Quantification: Automatic calculation of spores/mL and titer with dilution factors.
• 📈 Professional Export: Multi-sheet Excel reports with statistical analysis (Mean, SD, CV%).
• 📷 High-Res Support: Optimized for ArduCam 20MP and industrial microscopy cameras.
• 🎨 Modern UI: Auto dark/light mode (OS detection), HiDPI support, resizable panels, and real-time quality indicators.

---

Installation

Requirements

• Python 3.10 to 3.13
• A USB microscopy camera (ArduCam recommended)

Quick Install (PyPI)

pip install mira-microscopy
pip install opencv-python-headless --force-reinstall  # Linux only — avoids Qt conflict

Install from Source

git clone https://forge.ird.fr/phim/sravel/mira.git
cd mira
pip install .
pip install opencv-python-headless --force-reinstall  # Linux only

Quick Start

1. Launch MIRA: Run mira in your terminal.
2. Configure: Select your camera and load a YOLO .pt model.
3. Analyze: Select a chamber on the interactive grid, set your dilution factor, and capture images.
4. Export: Click "Save to Excel" to get your full report.

Pretrained Models

Four pretrained YOLO models accompany the MIRA paper (Mejias et al., 2026) and are distributed on the CIRAD Dataverse:

• Dataverse : https://dataverse.cirad.fr/dataverse/MIRA
• DOI : 10.18167/DVN1/E0JRDL

Model	Task	Weights (Dataverse)	mAP@0.5
M. oryzae spore counter	Detect (yolo26m)	spore-m-oryzae-xzewf_v9_yolo26m_20260721_0.765.pt	0.765
F. oxysporum micro/macroconidia	Detect (yolo26m)	spore-fusarium-object_v6_yolo26m_20260721_0.884.pt	0.884
P. fijiensis spore + surface	Segment (yolo26m-seg)	spore-fijiensis-seg_v6_yolo26m-seg_20260721_0.766.pt	0.770
Mix multi-genus (6 rice-associated)	Segment (yolo26m-seg)	curvularia-genus_v12_yolo26m-seg_20260721_0.820.pt	0.823

The Mix model covers 6 rice-associated fungal genera: Alternaria, Bipolaris, Curvularia, Exserohilum, Fusarium, Pyricularia.

Weights are distributed compressed with xz -9 (~12 % smaller). Decompress before use:

# Linux / Mac
xz -d *.pt.xz          # produces the .pt files

# Windows
# Right-click each .pt.xz → 7-Zip → Extract Here

Then place the .pt files in <repo>/models/, or set the environment variable:

export MIRA_MODELS_DIR=/path/to/models

Training your own models

MIRA models are trained with the VEGA CLI (Versatile Engine for Generic Automated-training):

https://forge.ird.fr/phim/sravel/vega

VEGA handles Roboflow dataset download, SLURM job submission, per-run reporting, and canonical weight naming. See its README for the full workflow.

Authors

Development Team:

• Sébastien RAVEL (CIRAD) - Lead Developer

  • Email: sebastien.ravel@cirad.fr
  • Role: Architecture, core features, camera integration

• Joffrey MEJIAS (CIRAD) - Co-Developer

  • Email: joffrey.mejias@cirad.fr
  • Role: UI/UX, detection algorithms, data export

Institution: CIRAD (Centre de coopération internationale en recherche agronomique pour le développement)

License

This project is licensed under the GNU General Public License v3.0 or later (GPLv3+).

Free for non-commercial use. For commercial licensing, contact the authors.

Citation

If you use MIRA in your research, please cite the software AND the accompanying dataset:

@software{mira2026,
  title     = {MIRA: Microscopy Image Recognition and Analysis},
  author    = {Ravel, S{\'e}bastien and Mejias, Joffrey and Adreit, Henri and Blanc, Ana{\"e}lle and Lubin, Nadia and Jolivet, Cassandre and Guyot, Valentin and Brayle, O{\"i}ana and Poncelet, Nicolas and Fournier, Elisabeth and Wicker, Emmanuel and Carlier, Jean and Tharreau, Didier},
  year      = {2026},
  institution = {CIRAD, UMR PHIM, Montpellier, France},
  url       = {https://forge.ird.fr/phim/sravel/mira},
  version   = {1.0.0}
}

@dataset{mira_dataverse_2026,
  title     = {MIRA: dataset and trained YOLO weights for automated counting and sizing of fungal spores --- France, 2024--2026},
  author    = {Ravel, S{\'e}bastien and Mejias, Joffrey and others},
  year      = {2026},
  publisher = {CIRAD Dataverse},
  doi       = {10.18167/DVN1/E0JRDL},
  url       = {https://doi.org/10.18167/DVN1/E0JRDL}
}

Support

For questions, bug reports, or feature requests:

• 📧 Email: sebastien.ravel@cirad.fr
• 🐛 Issues: https://forge.ird.fr/phim/sravel/mira/issues
• 📖 Docs: https://mira-microscopy.readthedocs.io

Made with ❤️ by CIRAD for the scientific community