AndroGen is an open-source synthetic data generation tool for automated sperm analysis from microscopic images. It allows researchers to generate highly customizable datasets without relying on real data or training generative models. Please check the web of the project for more details and examples.
- 📷 Synthetic generation of microscopic sperm images for multiple animal species.
- 🔧 Intuitive GUI with preloaded dataset configurations (SVIA, VISEM, BOSS).
- ⚙️ Full customization: morphology, concentration, movement, background, and debris.
- 🖥️ Compatible with both local and cloud environments via Gradio.
- 📂 Exports datasets with images, segmentation masks, and annotations (PNG, TXT).
- 🧪 Quantitatively and qualitatively validated using FID/KID metrics.
- 🧬 Modular and extensible: easily add new species, environments, motion patterns, and artifacts.
git clone https://github.com/dhernandezgit/AndroGen.git
cd AndroGen
./launch.shgit clone https://github.com/dhernandezgit/AndroGen.git
cd AndroGen
powershell .\launch.ps1And access this URL through your preferred explorer. From the GUI, you can:
- Load predefined configurations based on SVIA, VISEM, or BOSS datasets.
- Adjust visual parameters: background, brightness, contrast, blur, color, depth distribution.
- Define morphology: species, spermatozoa classes, dimensions, mobility, and more.
- Generate customized images or sequences and save them to your target directory.
AndroGen has been validated by replicating SVIA, VISEM, and BOSS datasets. Synthetic images were evaluated with:
- FID (Frechet Inception Distance)
- KID (Kernel Inception Distance)
Results show strong similarity to real datasets at low computational cost and without real data.
| SVIA Synthetic | VISEM Synthetic | BOSS Synthetic |
|---|---|---|
 |
 |
 |
AGPL-3.0 License: This open-source license is ideal for students and researchers, promoting open collaboration and knowledge sharing. See the LICENSE file for more details.
Enterprise License: Designed for commercial use, this license permits using this software for commercial solutions, bypassing the open-source requirements of AGPL-3.0. Please contact us for more details.
This work was supported by:
R&D project TED2021-129162B-C22, funded by MICIU/AEI/ 10.13039/501100011033/ and the European Union NextGenerationEU/ PRTR
R&D project PID2021-128362OB-I00, funded by MICIU/AEI/ 10.13039/501100011033/ and FEDER/UE.
- Daniel Hernández-Ferrándiz -
daniel.hernandezf@urjc.es - Juan J. Pantrigo -
juanjose.pantrigo@urjc.es - Soto Montalvo -
soto.montalvo@urjc.es - Raúl Cabido -
raul.cabido@urjc.es
Universidad Rey Juan Carlos, Móstoles, Spain