Melon Fruit Detection and Quality Assessment Using Generative AI-Based Image Data Augmentation

TL;DR

Limited large-scale labeled image datasets hinder melon-detection and quality assessment in agriculture. The authors employ generative AI (MidJourney and Firefly) to create three augmentation streams—text-to-image, image-to-image pre-harvest, and image-to-image post-harvest—and train/test YOLOv9 on real and AI-generated data. AI-generated images show substantial similarity to real images (PSNR roughly $27.4$–$28.8$, SSIM $0.06$–$0.42$), and YOLOv9 attains high detection performance (IoU around $0.95$) across conditions; net-quality metrics indicate realistic nets with some differences. This work demonstrates scalable data augmentation for agricultural vision tasks, enabling robust detection and external-quality assessment across crops.

Abstract

Monitoring and managing the growth and quality of fruits are very important tasks. To effectively train deep learning models like YOLO for real-time fruit detection, high-quality image datasets are essential. However, such datasets are often lacking in agriculture. Generative AI models can help create high-quality images. In this study, we used MidJourney and Firefly tools to generate images of melon greenhouses and post-harvest fruits through text-to-image, pre-harvest image-to-image, and post-harvest image-to-image methods. We evaluated these AIgenerated images using PSNR and SSIM metrics and tested the detection performance of the YOLOv9 model. We also assessed the net quality of real and generated fruits. Our results showed that generative AI could produce images very similar to real ones, especially for post-harvest fruits. The YOLOv9 model detected the generated images well, and the net quality was also measurable. This shows that generative AI can create realistic images useful for fruit detection and quality assessment, indicating its great potential in agriculture. This study highlights the potential of AI-generated images for data augmentation in melon fruit detection and quality assessment and envisions a positive future for generative AI applications in agriculture.

Figures (7)

• Figure 1: Overview of the experiment. Process of melon image data collection through cultivation experiments (A), three stages of image generation using generative AI (B), data augmentation and evaluation (C), YOLO model training (D), net quality assessment (E), and comprehensive evaluation.
• Figure 2: Text-to-image generation using Midjourney and Firefly (A), and the process of prompt generation from images followed by text-to-image generation (B)
• Figure 3: Image-to-image generation process using a greenhouse with melon plants before harvest as reference
• Figure 4: Image-to-image generation process using single melon fruit images after harvest as reference
• Figure 5: Data augmentation using generative AI across three image generation processes