IPB Summer Course Project: Advancing Smart Agriculture with Machine Learning
🔗 URL: About Course
⏱️ Working Period: 8–20 October 2023
Overview:
The International Summer Course of AI and Optimization for Smart Agriculture and Biomedical Applications held at IPB in October 2023 provided a platform for interdisciplinary collaboration and innovation in the field of data science and artificial intelligence (AI). As a Machine Learning Engineer, I spearheaded a project focused on clustering melon leaf images using advanced algorithms, contributing to the advancement of smart agriculture practices.
Project Details:
Objective: The primary objective of the project was to develop a clustering model for melon leaf images, leveraging the K-Means algorithm and gray-level co-occurrence matrix (GLCM) for feature extraction. By clustering the melon leaf data, we aimed to uncover patterns and similarities in leaf morphology, facilitating plant disease diagnosis and crop management strategies.
Methodology: We utilized the K-Means algorithm, a popular unsupervised learning technique, to partition the melon leaf images into distinct clusters based on their feature representations. Additionally, we employed the gray-level co-occurrence matrix (GLCM) to extract texture features from the leaf images, enhancing the discriminative power of the clustering model.
Implementation: The clustering model was developed using Python, leveraging libraries such as scikit-learn for machine learning algorithms and OpenCV for image processing tasks. We meticulously preprocessed the melon leaf images, extracted relevant features using GLCM, and applied the K-Means algorithm to perform clustering analysis.
After forming the clusters, we analyzed the resulting clusters by consulting with plant experts to understand the distinguishing features of each cluster. It was found that the clusters represented various conditions of the plants: some clusters indicated diseased plants, others showed nutrient deficiencies, some represented healthy plants, and a few clusters exhibited similar disease symptoms.
To determine the optimal number of clusters, we used Sum of Squared Errors (SSE) and silhouette score. After conducting several experiments, we found that the best clusters ranged between 3 to 6. Beyond this range, the differences between clusters were minimal. We chose to display 6 clusters, as it provided more comprehensive information when consulting with plant experts.
Achievements:
Our team’s exceptional performance in the IPB Summer Course Project garnered widespread recognition, including the following awards:
- Best Team: Our team was honored with the Best Team award for our collaborative efforts, innovative approach, and outstanding project outcomes.
- Best Foreign Participant: As a testament to our contributions and expertise, I was recognized as the Best Foreign Participant among the international cohort.
- Best Local Participant: Additionally, I received the accolade of Best Local Participant, highlighting my dedication, skills, and leadership in driving the project to success.
Impact:
The clustering model developed during the IPB Summer Course Project holds significant implications for smart agriculture and crop management practices. By accurately clustering melon leaf images based on their visual features, the model can assist farmers and agronomists in identifying potential disease outbreaks, optimizing pesticide application, and enhancing overall crop yield and quality.
Conclusion:
Participating in the IPB Summer Course Project was a rewarding experience that not only honed my technical skills but also provided invaluable opportunities for collaboration and knowledge exchange with international peers. Our success in developing an advanced clustering model for melon leaf images underscores the potential of AI and data science in revolutionizing agricultural practices for a sustainable future.
Screenshot :
Cluster | Cluster Plot |
---|---|
SSE Score | Silhoutte Score |
---|---|