Moth Scanner Demo

This notebook visualizes the results of trained models (detector and classifier) for the moth scanner presented in the following paper.

Installation

We suggest using miniconda. All required libraries are listed in the requirements.txt and can be installed with pip:

conda create -n scanner_demo python~=3.8.0
conda activate scanner_demo
pip install -r requirements.txt

Data and Model Weights

Data

The EU-Moths dataset can be downloaded from the given URL. In this notebook, we assume that the uncropped version of it is located under /home/korsch/Data/datasets/moths/eu_moths/uncropped/ORIGINAL. Set the root attribute of the Args class accordingly when you create an instance of it to match the location of your dataset. Please consult this GitHub repository to get more information about the MCC dataset.

Model Weights

Download the weights of the detector and the classifier and move them in the same directory as the jupyter notebook:

• Classifier (~ 84.7 MB)
• Detector (~ 78.8 MB)

Password for both downloads: Moth_Scanner

Licence

This work is licensed under a GNU Affero General Public License.