Entry No. 007
Junior (under 14)/ Senior (14 and above)
Senior (14 and above)
Category
6.2.1.1(6) Visionary Technologies
Classification
B. software
Country / Region Thailand
Name of Invention/Artwork PureRice: An Application for Inspecting Rice Milling Quality Using Instance Segmentation Techniques
Outline of the entry
The percentage of whole kernels is a critical indicator for assessing rice quality and determining market prices. However, the high cost and large size of conventional grading machines hinder their accessibility, particularly for small-scale farmers. To address this issue, the research team developed “PureRice,” a mobile application that leverages deep learning and computer vision technologies. Using the YOLOv8 model combined with instance segmentation techniques, the application analyzes rice images to accurately assess the percentage of whole kernels. The dataset, collected and enhanced through data augmentation, enabled the model to achieve a training accuracy of 94.8%. Application testing against samples from the Kalasin Rice Seed Center demonstrated a high level of accuracy, reducing error rates to only 5.9%. PureRice offers a low-cost, user-friendly solution for rice quality inspection, enhancing consumer confidence and supporting sustainable improvements in rice production. This study demonstrates the potential of integrating AI and computer vision to significantly improve the efficiency and accessibility of rice quality assessment and indicates strong possibilities for broader agricultural applications.
Characteristics of the entry
Rice is a vital economic crop for Thailand, serving as a major export commodity to several countries worldwide. In 2023, the value of Thailand’s rice exports reached USD 5.144 billion (approximately 178.136 billion baht), accounting for 19.2% of the country’s total agricultural exports (Office of Commerce, Chainat Province, 2024). Additionally, rice can be processed into various valuable products, such as skincare creams and rice noodles (Thongchai Suwannasin, 2017). The quality of milled rice is a key factor in price determination, typically assessed by the percentage of whole and broken kernels, based on national agricultural standards. However, current rice quality inspection remains challenging for farmers and rice traders, both large and small, as it often requires large-scale machinery or manual sorting, which can be costly and error-prone. Consequently, quality assessment practices are not widely adopted, resulting in market rice often failing to meet established standards (National Bureau of Agricultural Commodity and Food Standards, 2017). Recognizing these challenges, the researchers identified that advancements in artificial intelligence (AI) present a promising solution. By integrating AI into a mobile application, rice quality assessment can become more affordable, accessible, and efficient without relying on large, expensive equipment. This approach addresses the critical issues of cost and operational complexity, enhancing the accuracy and ease of rice quality inspections.
Demonstration details
The device serves to The rice detection app using AI starts by capturing rice images in a 15 × 15 cm frame from a 15 cm distance under controlled lighting. Images are checked for quality, then labeled into two classes: whole grain and broken rice using Roboflow, which also handles data augmentation e.g. rotation, cropping, brightness, noise.The model used is YOLOv8, trained on this dataset to detect and classify rice grains accurately. A JavaScript algorithm is developed to run the model and display results with bounding boxes.Finally, the model is integrated into a mobile app built with React Native Expo, allowing users to take or upload rice images and see detection results instantly and To begin using the application, users must first log in or sign in. Once successfully logged in, they can tap the camera button displayed on the screen to capture an image of the rice. The application will then analyze the image and assess the rice quality based on the detected characteristics.
Other notes about the entry (if any)
–
Information on patent, utility model, trademark, etc. application
–
Number of Team Members
2
Student
Miss Saruda Wannatong October 6, 2007 saruda2386@gmail.com
Master Aitthimon Bandasak June 28, 2007 krupea2@gmail.com