A customized convolutional neural network-based approach for weeds identification in cotton crops

Area of Science:

• Agricultural Science
• Computer Science
• Artificial Intelligence

Background:

• Soaring global food demand necessitates advancements in smart farming.
• Crop diseases and weed infestations significantly reduce agricultural yields.
• Cotton is a vital cash crop susceptible to various weeds and pests.

Purpose of the Study:

• To develop an efficient automated system for identifying and classifying cotton weeds.
• To compare the performance of a proposed deep CNN model against established CNN architectures.
• To address the challenge of managing large agricultural datasets for effective weed control.

Main Methods:

• Implementation of a deep convolutional neural network (CNN)-based architecture.
• Training and evaluation of the proposed model on cotton crop data.
• Comparative analysis with existing CNN models: VGG-16, ResNet, DenseNet, and XceptionNet.

Main Results:

• The proposed deep CNN model achieved a classification accuracy of 98.3%.
• VGG-16, ResNet-101, DenseNet-121, and XceptionNet achieved accuracies of 95.4%, 97.1%, 96.9%, and 96.1%, respectively.
• The proposed model demonstrated superior performance in identifying and classifying cotton weeds.

Conclusions:

• Deep CNNs offer a promising approach for automated weed identification in cotton cultivation.
• The developed model significantly improves upon existing methods for weed classification accuracy.
• Timely and efficient weed management through AI can mitigate yield losses in cotton farming.