Weed Recognition System In Potato Fields Based On DeepLearning ResNet-50 Model

This research explores the application of deep learning for automatic weed detection in potato fields, with particular considerations given to implementing the ResNet-50 Convolutional Neural Network model.Traditional weed management methods are labor intense, costly and environmentally damaging. This study seeks to improve precision agriculture practices by automating weed detection in order to reduce chemical herbicides reliance, increase agricultural efficiency, and in the process sustainability. Due to its capacity of feature extraction and image classification, the ResNet-50 model was selected as a suitable model for distinguishing crops and different weed species under difficult field conditions. For this, this study used a potato field dataset which was then collected, pre-processed and augmented to render a greater generalization for the model. First the metrics (accuracy, precision, recall, F1-score) on the performance of the model were evaluated. The results show reliable differentiation of weeds in potato fields with high accuracy, while it can facilitate targeted herbicide application and low impact on the environment. The findings from the study provide a scalable, cost effective solution to weed management on crop production that contributes to sustainable agriculture.