Detection of development stages of sunflower plant using visual transformers and comparison with other deep learning models

Throughout human history, several significant revolutions have taken place, chief among them being the Agricultural Revolution, the Industrial Revolution, and the Information and Digital Revolution, which emerged in the second half of the 20th century with the advent of computers, the internet, and digital technologies. These revolutions have fundamentally transformed societies, workplaces, economies, technologies, and cultures. Today, the agricultural sector faces global challenges such as population growth, climate change, and limited natural resources, necessitating the development of more efficient and sustainable production methods. Consequently, the integration of precision agriculture practices and artificial intelligence technologies plays a critical role in enhancing agricultural productivity and reducing environmental impacts. Particularly, accurately monitoring and analyzing the developmental stages of plants is crucial for optimizing resource use and improving product quality. This study focuses on tracking and analyzing the developmental stages of the sunflower plant. Within the scope of precision agriculture applications, data collected using a Canon EOS 700D from approximately 50 acres of land, consisting of 13 blocks of 4 acres each, at the Trakya Agricultural Research Institute's fields, were classified into seven categories based on their developmental stages. A total of 10502 data points were processed using artificial intelligence methods to obtain meaningful results. The study aimed to define and accurately model the different stages of sunflower development using modern AI techniques such as deep learning and convolutional neural networks, including VGG16, VGG19, InceptionResNetV2, and ViT models, alongside optimizers like SGD, AdaGrad, and Adam. Additionally, the potential impacts of advanced algorithms like transformer models on agricultural data analysis were explored. In the first part of our research, the theoretical foundations of precision agriculture and AI integration are detailed, and a literature review is provided to summarize existing knowledge on the subject. Subsequent sections offer comprehensive information on the developmental stages of the sunflower plant and explain the data collection processes in detail. The phases of training and testing deep learning models are then discussed, and the results are evaluated in terms of their implications for agricultural applications. This study aims to contribute to the development of agricultural technologies at both academic and practical levels. By emphasizing the importance of using AI and deep learning techniques to enhance productivity and support sustainable production methods in the agricultural sector, this research seeks to contribute to the digital transformation in agricultural production and serve as a resource for future studies.