Detection of Female Students’ Room Neatness Using YOLOv5

Abstract

Room neatness is an important aspect in creating a clean, comfortable, healthy, and supportive learning environment for female students in an Islamic boarding school. In the Al-Hasyimiyah area of Nurul Jadid, room neatness assessment is routinely carried out by the Environmental Conservation and Hygiene Division and regional administrators. However, manual assessment may lead to differences in perception among evaluators and requires considerable time and effort due to the large number of rooms to be assessed. This study aims to develop a female students’ room neatness detection system using the You Only Look Once version 5 (YOLOv5) method and implement it in a Flask-based web application. The dataset consisted of 1,000 room images divided into two classes: Neat Room and Untidy Room. The data were collected directly using a Canon EOS 1300D camera and then processed through resizing, annotation using Roboflow, and dataset splitting with a ratio of 70% training data, 20% validation data, and 10% testing data. The model was trained using YOLOv5s with an image size of 640 pixels, a batch size of 16, and 100 epochs. The training results achieved an mAP value of 0.989. Testing on 100 images showed that the Untidy Room class achieved a detection accuracy of 92%, the Neat Room class achieved 94%, and the overall accuracy reached 93%. The Flask web implementation allows users to upload room images and view detection results automatically. The findings indicate that YOLOv5 can be used as a supporting tool for assessing female students’ room neatness more quickly and objectively