A Deep Learning Approach for Classification of Cleanliness in Restrooms

Facilities hire cleaning companies to maintain and manage cleaning operations on their restrooms by deploying cleaners who are responsible for performing frequent checks to ensure the cleanliness of restrooms. Nevertheless, the perception of quality and word, clean is very subjective to the observer. Hence, it is not an easy task to quantify the cleanliness. This paper presents a deep learning approach using deep convolutional neural networks (DCNN) to detect and classify the level of cleanliness in restrooms into three different categories; namely dirty, average, and clean. Our method sheds new lights on data augmentation, feature extraction and knowledge transfer between models. The proposed architecture achieved a precision of 0.98, 0.95, 0.80 and recall of 0.99, 0.83, 0.95 for dirty, average, and clean categories respectively utilizing a dataset collected from an active restroom facility.