Image Segmentation for Crop-type Mapping

Corn and Soybean Mapping model at 30 m resolution from Landsat 8/ Sentinel 2 as presented in In-Season Wall-to-Wall Crop-Type Mapping Using Ensemble of Image Segmentation Models

Model Traininng

train/train.py - trains the model. Make sure the data directory/folder are set correctly. The training data can be exported as TFRecord from Earth Engine. The model expects data as 256x256 samples with the bands and the labels stacked together as channels. ( Refer to this notebook for reference)

Model Inference for Map Generation

generateMaps.py - to generate large maps from unlabeled data using a trained model.

Plotting Large Maps.

Refer to this Repo

Citation

If you find this code helpful, please cite the following work:

@ARTICLE{10323532,
author={Zaheer, Sheir A. and Ryu, Youngryel and Lee, Junghee and Zhong, Zilong and Lee, Kyungdo},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={In-Season Wall-to-Wall Crop-Type Mapping Using Ensemble of Image Segmentation Models},
year={2023},
volume={61},
number={},
pages={1-11},
doi={10.1109/TGRS.2023.3335214}}