Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data-Centric AI for Tropical Agricultural Mapping: Challenges, Strategies and Scalable Solutions

Mateus Pinto da Silva, Sabrina P. L. P. Correa, Hugo N. Oliveira, Ian M. Nunes, Jefersson A. dos Santos

TL;DR

The paper addresses the challenge of mapping agriculture in tropical regions with limited high-quality labeled data, cloud cover, and diverse crop calendars. It advocates a data-centric AI pipeline that prioritizes data quality and curation, rather than model complexity alone. It reviews 25 strategies across dataset creation, curation, training, and evaluation, and endorses 9 mature methods including confident learning, core-sets, specialized augmentation, and active learning. The work provides a practical pipeline for large-scale tropical mapping and signals a path toward robust, scalable, and generalizable agricultural mapping in tropical contexts.

Abstract

Mapping agriculture in tropical areas through remote sensing presents unique challenges, including the lack of high-quality annotated data, the elevated costs of labeling, data variability, and regional generalisation. This paper advocates a Data-Centric Artificial Intelligence (DCAI) perspective and pipeline, emphasizing data quality and curation as key drivers for model robustness and scalability. It reviews and prioritizes techniques such as confident learning, core-set selection, data augmentation, and active learning. The paper highlights the readiness and suitability of 25 distinct strategies in large-scale agricultural mapping pipelines. The tropical context is of high interest, since high cloudiness, diverse crop calendars, and limited datasets limit traditional model-centric approaches. This tutorial outlines practical solutions as a data-centric approach for curating and training AI models better suited to the dynamic realities of tropical agriculture. Finally, we propose a practical pipeline using the 9 most mature and straightforward methods that can be applied to a large-scale tropical agricultural mapping project.

Data-Centric AI for Tropical Agricultural Mapping: Challenges, Strategies and Scalable Solutions

TL;DR

The paper addresses the challenge of mapping agriculture in tropical regions with limited high-quality labeled data, cloud cover, and diverse crop calendars. It advocates a data-centric AI pipeline that prioritizes data quality and curation, rather than model complexity alone. It reviews 25 strategies across dataset creation, curation, training, and evaluation, and endorses 9 mature methods including confident learning, core-sets, specialized augmentation, and active learning. The work provides a practical pipeline for large-scale tropical mapping and signals a path toward robust, scalable, and generalizable agricultural mapping in tropical contexts.

Abstract

Mapping agriculture in tropical areas through remote sensing presents unique challenges, including the lack of high-quality annotated data, the elevated costs of labeling, data variability, and regional generalisation. This paper advocates a Data-Centric Artificial Intelligence (DCAI) perspective and pipeline, emphasizing data quality and curation as key drivers for model robustness and scalability. It reviews and prioritizes techniques such as confident learning, core-set selection, data augmentation, and active learning. The paper highlights the readiness and suitability of 25 distinct strategies in large-scale agricultural mapping pipelines. The tropical context is of high interest, since high cloudiness, diverse crop calendars, and limited datasets limit traditional model-centric approaches. This tutorial outlines practical solutions as a data-centric approach for curating and training AI models better suited to the dynamic realities of tropical agriculture. Finally, we propose a practical pipeline using the 9 most mature and straightforward methods that can be applied to a large-scale tropical agricultural mapping project.
Paper Structure (9 sections, 1 figure)

This paper contains 9 sections, 1 figure.

Table of Contents

  1. Introduction
  2. Mapping Tropical Agriculture
  3. Traditional Data Quality Assessment
  4. Data-Centric AI Approaches
  5. Creation of Datasets
  6. Data Curation
  7. Model Training
  8. Evaluation
  9. Final Remarks

Figures (1)

  • Figure 1: Data-Centric AI pipeline, adapted from Roscher et al.roscher2024better. DCAI approaches and their respective readiness in each step of the ML cycle are presented and contrasted with the traditional supervised ML pipeline. Letters A through D are further discussed in Sections \ref{['sec:dcai_creation']}, \ref{['sec:dcai_curation']}, \ref{['sec:dcai_training']}, \ref{['sec:dcai_evaluation']}, respectively. We chose not to include and cite methods with very low levels of readiness due to space constraints.