Year: 2022
Type: article
Abstract: Crop diseases have the potential to cause devastating epidemics that threaten the global food supply, and their dispersal patterns vary greatly. Therefore, it is essential to predict crop diseases to improve production efficiency hence, many existing techniques diagnose the disease through optical and automatic examination of infected leaves but they require additional feature extraction modules to predict only a single disease which results in time-consuming erroneous prediction. Hence an Ensemble ResNet-Efficient Net Model has been proposed that provides effective classification results on crop diseases with balancing network depth, width, and resolution thereby having limited computational resources, and precise and timely prediction of multiple diseases in crop images. Also, the features are extracted more significantly with the swish activation function in ensemble models without the need for a separate module which neglects erroneous prediction due to gradient vanishing and multi-co linearity. The experiment is conducted on 1950 real-time crop images collected from Crop fields using ResNet, Efficient Net B4, and Efficient Net B7 models. The result obtained shows that the proposed ensemble ResNet-Efficient Net model has high accuracy and low loss when implemented in Python and is compared to other existing methodologies.
Authors: Kundur, Niranjan C, Mallikarjuna, P B
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: diamond
OpenAlex ID: W7217364756