AI · Machine Learning
Melon Leaf Clustering
Unsupervised clustering of melon leaf images with K-Means and GLCM texture features — Best Team at the IPB International Summer Course.
- Role
- Machine Learning Engineer
- Timeline
- 8–20 Oct 2023
- Context
- IPB International Summer Course
- Stack
- Python
- scikit-learn
- OpenCV
- K-Means
- GLCM
Problem
At the International Summer Course on AI and Optimization for Smart Agriculture (IPB, Oct 2023), our team explored whether melon leaf images could be grouped automatically to support disease diagnosis and crop management.
Solution
- Extracted texture features with the gray-level co-occurrence matrix (GLCM).
- Clustered images with K-Means (scikit-learn, OpenCV for pre-processing).
- Chose the number of clusters with SSE and silhouette score — useful values ranged 3–6; we used 6 because it gave the most information to plant experts.
- Reviewed clusters with plant experts: they mapped to diseased leaves, nutrient deficiencies, healthy plants, and similar-symptom groups.
Impact
- clusters validated with plant experts
- 6clusters validated with plant experts
- award
- Best Teamaward
Our team won Best Team, and I received the Best Local Participant award.