Sistema para detecção de problemas na cultura do morango a partir da análise de imagens

Abstract

This study aims to leverage machine learning techniques in developing a model for classifying images within the realm of agriculture. We focus on a dataset tailored to strawberry plants, a common crop found in small-scale farming. The model was implemented using the Ultralytics HUB analysis app, renowned for real-time object detection and image recognition, and it’s fine-tuned with the power of GPUs. Our approach involves utilizing a Kaggle platform dataset, transferred seamlessly to RoboFlow, and adopting the YOLOv8 architecture. The primary goal is to explore methods for identifying plant issues, such as damage or diseases, thereby minimizing the need for manual inspections. By automating the image verification process, we envision the ability to monitor a larger number of cultivated areas, generate more precise reports, and reallocate resources to address other vital agricultural needs. This contributes to the efficient use of resources available to small-scale farms. The final result obtained was an application that, when connected to the internet, can make predictions about strawberry diseases (without explaining the causes or explaining the reasons for the diseases) when the camera is pointed at them.