balloon-data_analysis

codes for analyzing B-ICI balloon data

Use typhon package for ginding equilibrium vapor pressure of water and ice https://www.radiativetransfer.org/misc/typhon/doc/index.html

Convert the PTU and the ETAG file from esrange. Use 'plots.py' to create the plot, and 'ptu_data.py' to convert the ptu file. The ETAG file might have to be tailored bz each launch. The code expects a specific form of the code, and that can varies with each launch. Check if the format looks exactly like how the program expects it.

Using the models

In order to use the models, create a virtual environment (with conda).

Segmentation model

Clone the git-repo to ".../data_analysis_tutorial/."

https://github.com/Omnitok/snow_crystal_segmentation

Use the dockerfile for installing the dependencies "*/data_analysis_tutorial/snow_crystal_segmentation/model/docker" "docker build -t cuda-tensorflow ."

The model is in snow_crystal_segmentation folder. Already pretrained, currently using model #m232.

Run it with "sudo ./run_inference.sh". Here you can change the input-output directories.

On "run_inference.py" one can set the detection limit for the model, and also can tweak the particle properties that the model will produce.

Classification model

python=3.9 install the dependencies with "pip install -r requirements.txt" The model is in the folder "classification". Pretrained, but ideally with every measurements the hand/automaticly classified particles are fed back to the training, to train a new model.

Run it with "classify.py" where the input/oputput folders can be set.