Melon Fruit Detection and High quality Evaluation Utilizing Generative AI-Primarily based Picture Information Augmentation
Authors: Seungri Yoon, Yunseong Cho, Tae In Ahn
Summary: Monitoring and managing the expansion and high quality of fruits are essential duties. To successfully prepare deep studying fashions like YOLO for real-time fruit detection, high-quality picture datasets are important. Nonetheless, such datasets are sometimes missing in agriculture. Generative AI fashions may also help create high-quality photos. On this examine, we used MidJourney and Firefly instruments to generate photos of melon greenhouses and post-harvest fruits by text-to-image, pre-harvest image-to-image, and post-harvest image-to-image strategies. We evaluated these AIgenerated photos utilizing PSNR and SSIM metrics and examined the detection efficiency of the YOLOv9 mannequin. We additionally assessed the web high quality of actual and generated fruits. Our outcomes confirmed that generative AI might produce photos similar to actual ones, particularly for post-harvest fruits. The YOLOv9 mannequin detected the generated photos properly, and the web high quality was additionally measurable. This reveals that generative AI can create sensible photos helpful for fruit detection and high quality evaluation, indicating its nice potential in agriculture. This examine highlights the potential of AI-generated photos for knowledge augmentation in melon fruit detection and high quality evaluation and envisions a optimistic future for generative AI purposes in agriculture.