Publication: Design and implementation of lightweight model for agricultural crop recognition
Loading...
Date
2024
Authors
Harivindran a/l Ramalingam
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Crop recognition is essential for enhancing agricultural productivity and efficiency. Traditional methods often involve manual efforts and subjective judgment, posing significant challenges. To address these issues, this project proposes the design and implementation of a lightweight model for agricultural crop recognition using the EfficientNet-B0 architecture, chosen for its speed, simplicity, and scalability. EfficientNet-B0 is preferred over heavyweight models like ResNet due to its optimized performance with fewer parameters and lower computational costs. Heavyweight models, while accurate, require substantial computational and memory resources, making them impractical for resource-constrained environments like agricultural fields. EfficientNet-B0's compound scaling method provides high accuracy and efficiency suitable for real-time deployment in resource-limited settings. The project involves collecting datasets, training models, and evaluating performance. Notably, EfficientNet B0 achieved an accuracy of 99.20% at 60 epochs, with a precision of 100%, recall of 100%, and an F1-score of 100%. In comparison, ResNet achieved an accuracy of 97.86%, with a precision of 100%, recall of 100%, and an F1-score of 100%. This demonstrates the effectiveness of lightweight models in resource-constrained environments. Ultimately, this approach aims to enhance agricultural technology and practices by improving the efficiency, accuracy, and accessibility of crop recognition.