Improve the Deep Learning Models in Forestry Based on Explanations and Expertise

Code and data for the article "Improve the Deep Learning Models in Forestry Based on Explanations and Expertise" by Ximeng Cheng, Ali Doosthosseini and Julian Kunkel (2022). (https://doi.org/10.3389/fpls.2022.902105)

Requirements

Data

The PlantVillage dataset used in the article is publicly available. It can be found at: https://github.com/spMohanty/PlantVillage-Dataset.

This dataset should be downloaded and placed inside data/

GradCAM

The explanations require GradCAM to be installed. See https://github.com/jacobgil/pytorch-grad-cam.

Usage

1. Split the data into train/test/validation sets using the scripts in utils. Set the masks as desired.
2. Build and train the model using the methods in utils/model.py. For the explanations, use the methods in utils/explanables.py.

See experiment_1.py, experiment_2.py,experiment_3.py for the experiments used in the article.

Citation

Please consider citing our paper if it helps in your work:

Ximeng Cheng, Ali Doosthosseini, & Julian Kunkel (2022). Improve the Deep Learning Models in Forestry Based on Explanations and Expertise. Frontiers in Plant Science. 13:902105. DOI