Enhancing Sugarcane Disease Diagnosis: UNET Architecture for Red Rot Identification

Image segmentation is an important part of many computer vision problems and existing approaches such as pixel by pixel classification and manual annotation can be either in efficient or inaccurate. This research explored the way UNET architecture can circumvent these constraints to yield accurate segmentation. A diverse image dataset was applied to the UNET, and then its performance compared to conventional segmentation methods. Its capability to capture both the local and global characteristics contributes to the model’s improved performance, especially in complex settings. To detect red rot disease in sugarcane, these characteristics were used with an accuracy of 97%. UNET can be further improved by integrating it with other deep-learning techniques and its future research should focus on optimizing UNET for special use cases. This technology can also be used by farmers and agricultural experts to better pinpoint and tackle crop diseases, which could lead to better harvest yields and less financial baggage.