Rekibai: A Cnn-Based Mobile Application For Real-Time Household Waste Classification

The accelerated growth of urban solid waste has intensified the need for technological tools that support correct separation of household waste. This study presents REKIBAI, a mobile application designed to assist users in the classification of domestic waste using computer vision techniques. The system integrates a ResNet-50-based convolutional neural network for image classification, deployed within a mobile architecture connected to cloud-based services for data storage and processing. The model was trained using a dataset of 10,464 labeled images distributed in six waste categories (plastic, paper, metal, biodegradable, cardboard and glass), applying a 70/20/10 split for training, validation and testing. Experimental results indicate an overall classification precision of 87% for specialized materials. The system was evaluated with two user groups under ISO 9241-11 usability criteria. The application achieved a 90% general acceptance rate and a System Usability Scale (SUS) score of 80/100, reflecting high perceived usability. Comparative analysis showed consistent performance across user profiles, with minor differences in efficiency and error rates. The findings demonstrate that REKIBAI constitutes a viable technological alternative to promote responsible waste management at the household level, combining real-time image recognition, mobile deployment and user-centred design principles.