This folder contains Jupyter notebooks to help you train your own models and analyse your measurements.
The scripts have the following dependencies and were tested with the following versions:
- tensorflow: 1.13.1
- keras: 2.3.1
- seaborn: 0.10.0
- eli5: 0.10.1
- pandas: 1.0.3
StarDist provide examples to train their model here.
Our model was trained on 100 epochs, with 48 rays, images of 320x320 pixels and with a dropout of 0.5. With GPU enabled it should take about 30 minutes to train on Google Colab.