International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 04 | Apr 2025
p-ISSN: 2395-0072
www.irjet.net
Optimized Fruit Quality Prediction using CV and GANs Ashish B Rao1, Dr. Tejaswini R Murgod 2, Loukya Harisha3 Ananya Nittur4 1,3,4Student, Department of AIML, BNM Institute of Technology, Bangalore, Karnataka, India 2Professor, Department of AIML, BNM Institute of Technology, Bangalore, Karnataka, India
----------------------------------------------------------------------***--------------------------------------------------------------------estimation and simulation of temporal degradation, the Abstract - A modern approach aimed at transforming supply
system has the ability to forecast not only the current state of quality but also future possible shelf-life trajectories. Through the application of image-based quality estimation along with predictive shelf-life modeling, the system offers practical insights to farmers to facilitate well-informed decisions about harvesting, storage, and distribution, thus achieving optimal supply chain management.
chain and agricultural operations is Optimized Fruit Quality Prediction using GANs and Computer Vision. This system offers precise fruit quality evaluations by combining Generative Adversarial Networks with cutting-edge computer vision, making sure of increased market prices due to better quality and reducing harvest processing losses. An accurate CNNbased fruit quality classifier, a GAN-based synthetic data augmentation model, and a natural graphical user interface (GUI) for looking at quality indicators and predictions are some of the key features. These key factors improve the agricultural ecosystem's sustainability and productivity by giving stakeholders practical insights. Key Words — Fruit Quality Prediction, GANs, Computer Vision, Agricultural Technology, Data Augmentation, CNN, Supply Chain Efficiency.
The method of interest has its roots in measuring banana ripeness as a demonstrative example of how accurate technological measurement can coexist with laboratory precision to actual usage for agriculture. With this system, the goal is to establish a new benchmark for functional, AI-driven quality control that unites scientific excellence with operational feasibility for a broad scope of agricultural stakeholders.
1. INTRODUCTION
2. LITERATURE REVIEW
Fruit quality evaluation and shelf-life prediction remain major issues for agri-food supply chains, with direct consequences for farmers' profitability and minimization of food wastage. The use of subjective, time-consuming, and variable inspection procedures results in the early withdrawal of edible products or delayed detection of spoilage, causing severe economic losses and sustainability problems along the value chain. For perishable fruits like bananas, where the ripening process plays a major role in their market value, even minor mistakes in assessing quality can lead to significant financial losses. In fact, post-harvest losses in developing countries are estimated to range between 30% and 40%.[8]
Deep learning has been extensively explored for improving agricultural processes, particularly in small-scale agriculture. [1][5][10] For instance, V. Zárate and D. C. Hernández (2024) highlight the use of lightweight deep models to assess fruit quality within small-scale agricultural settings with limited computational resources. Scalability remains a problem owing to the constraints of lightweight models. Similarly, M. Zabovnik and D. Wojcieszak (2024) report on the applications of Convolutional Neural Networks (CNNs) in smart agriculture, which include multispectral images and vegetation indices, while their study is predominantly a review without experimental verification. J. Li and P. Kumar (2024) introduced ensemble deep learning architectures for cotton crop classification in AI-based smart agriculture that enhanced the decision-making process through image analysis. Their method does use significant quantities of computational power, which could restrict it from being used in resource-limited environments.
Existing technologies suffer a compromise between f ield use and accuracy,while heterogeneity in environments a nd insufficient data destabilize systems that are computerize d.Traditional methods based on manual sorting or simple computer vision techniques are incapable of identifying the complex biochemical processes in fruit maturation, while laboratory-grade instruments are not feasible for implementation in agricultural environments.[10] These issues highlight the need for efficient, scalable solutions that can be implemented in real-world environments with low infrastructural requirements. This article proposes a technology-based solution for such issues utilizing computer vision and deep learning techniques. By utilizing a combination of multi-level quality
© 2025, IRJET
|
Impact Factor value: 8.315
|
There have also been some recent advances in deep learning that have involved transfer learning and image augmentation. E. Martinez and F. Lopez (2024) utilize transfer learning to classify fruits such that the training time taken was reduced while the accuracy was increased. However, domain adaptation remains a problem, affecting the model's generalizability to different types of fruits and conditions. In addition, image augmentation techniques have been explored for robust fruit detection using deep learning. While these techniques improve model robustness and
ISO 9001:2008 Certified Journal
|
Page 1951