UPLB ICS Peak One

This study investigates the feasibility of using single- stage neural networks, particularly YOLOv7, for automating game recording in over-the-board (OTB) chess tournaments. Traditional manual recording on scoresheets is time-consuming, prompting interest in digital solutions. However, existing dig- ital chessboards are costly. The trained model achieved an mAP@.5:.95 of 0.6328, demonstrating effectiveness close to the state-of-the-art. While it exhibited high accuracy in detecting most pieces, challenges arose with pieces of similar appearances and those positioned farther from the camera. Portable Game Notation (PGN) generation yielded 100% accuracy for standard chess sets, yet smaller sets necessitated adequate lighting to enhance detail for enhanced accuracy.

One of the major problems that education would like to address is the facilitation of learner’s engagement and motivation toward learning. Further challenged by the COVID- 19 pandemic and the boom of numerous technological advances that disrupted the orthodox and unorthodox learning experi- ence. ÆDOM is a web platform that aims to utilize emerging technological paradigms such as the Metaverse and Blockchain to address and provide a reinforced learning environment. This study presents the development, implementation, and assessment of ÆDOM in delivering such service to Senior High School. The application was evaluated using methods such as System Usability Scale (SUS), Intrinsic Motivation Inventory (IMI), and data analytics within the system. It is concluded that the application is highly acceptable and usable while engaging learners’ motivation towards learning as defined in the Self-Determination Theory as its behavioral model. Results from this paper contribute to similar fields of study.

The Philippines is a disaster-prone country that needs to implement technologies to ensure the safety of its people. As such, this study aims to create a community-powered geographic information system in order to monitor ongoing disasters, and classify their severity in order to aid in disaster response. The source of data used were tweets from Twitter. The web program was able to pull data, classify tweets, and display in a Google Map. Further research can be done on two ways, by using other text classifiers besides Naive Bayes, or by implementing other ways of finding location from text such as Named Entity Recognition.

There are countless news articles online that people have many different ways of consuming them (news aggregators, etc.). This study proposes a mobile application that informs the user of information regarding crime, accidents, and natural disaster occurrences of their current location in an easy-to- digest manner– through maps and push notifications. The process includes scraping articles online, categorizing and extracting information using a a finetuned DistilBERT model, storing to a database then displaying on the mobile application. The web scraper and DistilBERT model had an accuracy percentage of -250%, 78.57%, and 100% for natural disaster, crime, and accident respectively. The mobile application had a System Usability Scale score of 81.25 evaluated to ”usable and can be recommended.”

Expresscript is a mobile application designed for healthcare providers and patients that utilizes current features and technology of mobile devices in enhancing their medical treatment course. The application allows healthcare providers to generate medical e-prescriptions that lessens the chances of medication inaccuracies and errors. The application also allows patients to automatically generate schedules from medical e- prescriptions that will aid them in complying with their medical regimen. The usability of the applcation was evaluated using the System Usability Scane (SUS) with a score of 72.75 which can be classified as above average. Feedback and suggestions from respondents showed great interest for the application.