Citation: Salamai, A.A.;
Al-Nami, W.T. Sustainable Coffee
Leaf Diagnosis: A Deep
Knowledgeable Meta-Learning
Approach. Sustainability 2023, 15,
16791. https://doi.org/10.3390/
su152416791
Academic Editor: Ripon Kumar
Chakrabortty
Received: 21 August 2023
Revised: 28 September 2023
Accepted: 28 September 2023
Published: 13 December 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
sustainability
Article
Sustainable Coffee Leaf Diagnosis: A Deep Knowledgeable
Meta-Learning Approach
Abdullah Ali Salamai
1,
* and Waleed Tawfiq Al-Nami
2
1
Department of Management, Applied College, Jazan University, Jazan 82822, Saudi Arabia
2
Department of Computer, Applied College, Jazan University, Jazan 82822, Saudi Arabia;
walnami@jazanu.edu.sa
* Correspondence: absalamai@jazanu.edu.sa
Abstract: Multi-task visual recognition plays a pivotal role in addressing the composite challenges en-
countered during the monitoring of crop health, pest infestations, and disease outbreaks in precision
agriculture. Machine learning approaches have been revolutionizing the diagnosis of plant disease in
recent years; however, they require a large amount of training data and suffer from limited general-
izability for unseen data. This work introduces a novel knowledgeable meta-learning framework
for the few-shot multi-task diagnosis of biotic stress in coffee leaves. A mixed vision transformer
(MVT) learner is presented to generate mixed contextual attention maps from discriminatory latent
representations between support and query images to give more emphasis to the biotic stress lesions
in coffee leaves. Then, a knowledge distillation strategy is introduced to avoid disastrous forgetting
phenomena during inner-loop training. An adaptive meta-training rule is designed to automatically
update the parameters of the meta-learner according to the current task. The competitive results
from exhaustive experimentations on public datasets demonstrate the superior performance of our
approach over the traditional methods. This is not only restricted to enhancing the accuracy and
efficiency of coffee leaf disease diagnosis but also contributes to reducing the environmental footprint
through optimizing resource utilization and minimizing the need for chemical treatments, hence
aligning with broader sustainability goals in agriculture.
Keywords: sustainable; coffee leaf diseases; leaf diagnosis; disease management; artificial intelligence;
meta-learning
1. Introduction
Multi-task visual recognition is a revolutionary computer vision paradigm that allows
training single artificial intelligence algorithms to perform many visual recognition tasks
instantaneously. Different from single-task learning, in which independent models are
trained for each separate task, resulting in redundant computation and enlarged complexity
of the recognition algorithm, the multi-task paradigm makes use of shared representations
across tasks, enabling the model to learn common designs and representations that are
valuable for multiple tasks [1]. This method encourages the model to learn from its
experiences with different tasks, thus enhancing its performance across the board. The
model’s efficiency and efficacy are both improved by the shared learning across tasks,
and more comprehensive comprehension of the visual input is provided, allowing for the
modeling of complex interdependencies and linkages. Better generalization is another
benefit of multi-task visual recognition since the model is trained to extract high-level
features that are robust and meaningful across different recognition tasks [2].
Deep learning (DL) has been revolutionizing the visual recognition tasks
(i.e., classification, detection, segmentation, colorization, etc.) in many application domains
owing to the powerful learning capabilities that enable the DL models to automatically
process and extract patterns of disease from plant images without the need for any hand-
crafted features or engineered features [3,4]. Among the DL methods, convolutional neural
Sustainability 2023, 15, 16791. https://doi.org/10.3390/su152416791 https://www.mdpi.com/journal/sustainability