Download the segmentation model at https://doi.org/10.82296/hmgu-nefeli.2yphm-15n14 and save it as displayed in the required folder structure.
This code was tested on an M1 chip and a conda environment. For using the GPUs of macOS, please uncomment the macOS requirements in requirements.txt.
conda create -n echovisuall python=3.11
source activate echovisuall
pip install -r requirements.txt
For running on NVIDIA GPUs, we provide a Dockerfile that creates a Docker Image. Make sure to copy the segmentation model as described in the required folder structure. Follow these instructions for running with Docker and Docker Compose:
git clone https://github.com/ExperimentalGenetics/EchoVisuALL.git
docker compose up --build
| Folder | Content | Description |
|---|---|---|
| data | Example data | |
| notebooks | Folder with demo notebooks | |
| results | parameters | Storage locations for results |
| predictions | ||
| segmentation_model | variables | Download the segmentation model and insert here |
| fingerprint.pb | ||
| keras_metadata.pb | ||
| saved_model.pb | ||
| src | Sources folder | |
| main.py | Main entry point | |
| Dockerfile | Dockerfile | |
| docker-compose.yml | File used by Docker Compose | |
| requirements.txt | Installation requirements | |
| README.md |
You can run this workflow by using main.py or the provided notebooks.
| Workflow | Description |
|---|---|
| main.py | This Python file can be used for an end-to-end analysis by calling the main method. |
| end2end_workflow.ipynb | This notebook can be used for an end-to-end analysis. |
| stepwise_workflow.ipynb | This notebook can be used for a stepwise analysis. |
Change these settings in main.py or within the notebooks to change behaviour.
| Parameter | Description |
|---|---|
| mc_samples | This parameter is set to determine the number of Monte Carlo samples used by the Bayesian Network to provide segmentation masks and the confidence measure. Originally, it was set to 10. Larger values lead to an increase of computational time. |
| save | This parameters must be set to True, if results should be stored automatically. |
| iso | This parameter, determines the filtering threshold that depends on the anaesthetic of the mouse. It is set to True, by default. If you like to use the workflow on data from awake mice, please set iso=False. |