Weak Labeling for Cropland Mapping in Africa
A new method improves Africa’s cropland mapping by refining weak labels through unsupervised clustering. Combined with minimal human annotations, it greatly boosts segmentation accuracy—raising the cropland F1 score from 0.53 to 0.84 with only 33 labeled samples—helping overcome scalability limits in high‑resolution mapping.
Subject Tags
- Data Science and Artificial Intelligence
- Agriculture
- Conservation Technology
Abstract
Cropland mapping can play a vital role in addressing environmental, agricultural, and food security challenges. However, in the context of Africa, practical applications are often hindered by the limited availability of high-resolution cropland maps. Such maps typically require extensive human labeling, thereby creating a scalability bottleneck. To address this, we propose an approach that utilizes unsupervised object clustering to refine existing weak labels, such as those obtained from global cropland maps. The refined labels, in conjunction with sparse human annotations, serve as training data for a semantic segmentation network designed to identify cropland areas. We conduct experiments to demonstrate the benefits of the improved weak labels generated by our method. In a scenario where we train our model with only 33 human-annotated labels, the F1 score for the cropland category increases from 0.53 to 0.84 when we add the mined negative labels.
Citation
Hacheme, G.Q., Zaytar, A., Tadesse, G.A., Robinson, C., Dodhia, R., Ferres, J.M.L. and Wood, S., 2024, July. Weak labeling for cropland mapping in africa. In IGARSS 2024-2024 IEEE International Geoscience and Remote Sensing Symposium (pp. 258-262). IEEE. 10.1109/IGARSS53475.2024.10640756
TNC Authors
-
Stephen Wood
Applied Scientist
The Nature Conservancy
Email: stephen.wood@tnc.org