Published Paper Details:

Weeds in rice fields are a major problem because they compete with rice plants for nutrients, light and space, resulting in reduced yields and economic losses. Physical detection of weeds is problematic when human labor is scarce. The proposed solution includes a deep learning approach that uses a specially pre-trained DenseNet-121 model to increase efficiency in solving the problem of the negative impact of weed growth on rice crop by distinguishing between visually similar weeds and crops. This model is refined with a dataset created from images of rice fields, and data augmentation techniques are used to improve the reliability of the model. This approach implements and evaluates various metrics and improves accuracy by displaying test results. This method helps in accurate weed control which ultimately improves yield and reduces economic losses.The system underwent evaluation using a dataset of images sourced from a paddy field, demonstrating its capability of identifying and eliminating weeds with a precision exceeding 99%. The suggested approach has the potential to greatly minimize labor and environmental impacts associated with weed management, while enhancing the efficiency and precision of weed detection and removal in crop production.. In future integrating this model with real-time observation systems and applied knowledge will further improve its practical applicability in agriculture.