A robotic assistant for pictogram classification in education: a proposal using dynamically quantized deep neural networks and the CIFAR-10 dataset for developing countries

Currently, there exists a wide range of deep learning models developed for numerous tasks, ranging from automatic speech recognition to music and video generation. According to various authors, these models hold significant potential to contribute to achieving several Sustainable Development Goals (SDGs) established by the United Nations. However, in developing countries such as Ecuador, not all educational institutions-particularly those in rural areas- have access to the necessary infrastructure to implement these models in ways that enhance educational processes for children. In response to this issue, this study presents a low-cost robotic assistant that utilizes quantized deep learning networks to support the recognition of pictograms in basic general education. The proposed system was tested with a group of 52 children between the ages of 5 and 8, yielding a Cronbach’s Alpha coefficient of 0.71, which suggests that the solution is promising.