This project uses image data to distinguish desired crops from unwanted weeds. The classifier used is ResNet18 and the data comes from the Sugar Beets 2016 Dataset. This repository contains the scripts necessary to preprocess the dataset and the notebooks necessary to train the classifier. You will need to prepare the dataset yourself, but the notebooks can be run in Colab:
The classifier is far from perfect, but this project helped me learn a lot about data processing, unevenly distributed data sets, and PyTorch. Performance metrics can be viewed in the reports and presentations.
- scripts: python files for preprocessing the input dataset
- data: folder where the data resides for the preprocessing functions (omitted by
.gitignore, you will need to create your own) - pytorch-vision: open source helper functions
- img: supporting images
- reports+slides: final report and slides for EE244
This project uses images from the Sugar Beets 2016 Dataset. The data can be downloaded here. Please cite their work! :)
@article{chebrolu2017ijrr,
title = {Agricultural robot dataset for plant classification, localization and mapping on sugar beet fields},
author = {Nived Chebrolu and Philipp Lottes and Alexander Schaefer and Wera Winterhalter and Wolfram Burgard and Cyrill Stachniss},
journal = {The International Journal of Robotics Research},
year = {2017}
doi = {10.1177/0278364917720510},
}