Helmets Labeling Crops: Kenya

Code for paper: Helmets Labeling Crops: Kenya Crop Type Dataset Created via Helmet-Mounted Cameras and Deep Learning

Crop Type Dataset

Pipeline

The pipeline of generating a crop type dataset involves the following steps:

1. GoPros are used for collecting road side images in agricultural zones (Helmets Data Collection Guide)
2. Collected photos are uploaded to Cloud Storage (Instructions for Uploading)
3. Photos are processed using our ML models and turned into a Google Earth Pro KMZ file (notebooks/1_GoPro2CropKMZ.ipynb)
4. KMZ file(s) are analyzed to verify model predictions (Helmets Quality Assessment Instructions)
5. Reviewed KMZ files are converted into a csv file (notebooks/2_CropKMZtoCSV.ipynb)
6. CSV files are combined into a single Kenya dataset (notebooks/3_Kenya_dataset_publish.ipynb)

Machine Learning Models

Scripts for training and validating the CropNop and CropSeg model is in the ML folder.

Working with the points

Google Earth Engine repository

1. Crop type points visualized in GEE: gee/1_all_checked_points.js
2. Crop type points used for mapping in GEE: gee/1_all_checked_points.js