Enhancing Maize Seed Quality: Defect Detection using Google Net.

Information Technology Engineering

This study explores the application of deep learning techniques, specifically Google Net, to detect maize seed defects. The research uses convolution neural networks (CNNs) and transfer learning to accurately identify and classify defects in maize seeds. The process begins with dataset curation and preprocessing, followed by the adaptation of the Google Net architecture. The model is trained, validated, and tested, providing insights into its efficacy. Maize, a staple crop worldwide, requires meticulous seed quality control to ensure optimal yield. Traditional methods of seed inspection are labor-intensive and susceptible to human error. Deep learning techniques offer a promising avenue for automating seed defect detection. Google Net, with its intricate architecture and remarkable feature extraction capabilities, emerges as a potent tool for accurate and efficient defect identification. The study collects a diverse dataset containing images of maize seeds with various defects, labeled to encompass fungal infections, physical damage, and discoloration. Data preprocessing involves resizing images and applying data augmentation techniques to enhance model generalization. The Google Net architecture, consisting of parallel convolution pathways and filters of varying sizes, excels in capturing intricate patterns and features. The model is initialized with pre-trained weights from Image Net, fine-tuned for the maize seed defect detection task, and customized for effective distinction between healthy seeds and those exhibiting defects. The model's efficacy is evaluated on an independent testing dataset, providing a comprehensive view of its defect detection capabilities. In conclusion, this study demonstrates the viability of using Google Net for maize seed defect detection, capitalizing on the strengths of CNNs and transfer learning. This technology offers an automated and efficient solution to seed quality control, contributing significantly to crop production and global food security.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.