CME Journal of Clinical Case Reports
ISSN: 3067-7998 | DOI: 10.64030/3067-7998
Binshad M S and Greeshma K V
Cannabis seed classification is essential for precision agriculture, as it supports genetic integrity, regulatory compliance, and crop productivity. Manual identification methods are often slow and prone to error, creating demand for automated solutions. This study introduces a machine learning approach for cannabis seed detection and classification using a curated dataset of 3,434 seed images across 17 varieties. Images were standardized through resizing and grayscale conversion, and pixel intensity values were extracted to form feature vectors. A Support Vector Machine (SVM) classifier was trained and evaluated, achieving an overall accuracy of 93.98%. Several varieties, including AK47_photo, Hang Kra Rog KU, and Thaistick Foi Thong, were classified with perfect precision and recall. The macro-average F1-score of 0.93 and weighted-average F1-score of 0.94 confirm balanced performance across categories. These findings demonstrate the effectiveness of SVM-based grayscale modeling for scalable cannabis seed classification. Future work will focus on expanding dataset diversity and integrating advanced feature extraction and deep learning methods to further enhance model robustness.