Evaluation scripts for instance segmentation

currently under construction

Developed for instances segmentation in 2d and 3d. Computes a number of evaluation metrics (e.g., AP, F1, coverage, precision, recall). Provides visualization of the errors.

installation:

The recommended way is to install it into your conda/python virtual environment.

conda activate <<your-env-name>>
git clone https://github.com/Kainmueller-Lab/evaluate-instance-segmentation
cd evaluate-instance-segmentation
pip install -e .

usage:

It is currently use by importing the module and calling the evaluate_file function. It will be updated to be callable from the command line.

output:

• evaluation metrics are written to toml-file and returned as dict