Enhanced Image Classification Quality using Diffusion Synthesis: a Case Study in Sugar Industry

The classification of sugarcane quality by computer vision is important for optimizing sugar production. This research focuses on improving a system for sugarcane quality assessment. The system uses deep learning models, including MobileNetV2 and ResNet50V2, and incorporating synthetic data generation through diffusion models. The synthetic data, generated through a controlled blending process and segmentation-base augmentation, is used to enhance the system performance. The result of the study shows a classification accuracy of 98% across five sugarcane quality categories.