Deep Learning: Traffic Accident Captioning Model in Madagascar Mother Language
Soniarimamy Nantenaina Serge Rochel, Razafindramintsa Jean Luc, Thomas Mahatody, Manantsoa Victor · 2022 8th International Conference on Control, Decision and Information Technologies (CoDIT) · 2022
This paper presents a deep learning model for road accident captioning in the Malagasy language. As we have observed, the road accident is a deserved study case due to the relentless increase in Madagascar. In addition, image description in the Malagasy language is a subject of research that is not yet realized. Thus, we have created our dataset in the Malagasy language to do this. We have also adopted deep learning and word embedding as a method. As a result, we obtained a deep learning model which can describe a road accident image in the Malagasy language at 95.9898 percent of accuracy. The results obtained in this research are beneficial to protecting the life of Malagasy people in a road accidents. Moreover, our research is distinguished from all the related works in this domain by collaborating with an expert in the Malagasy language during the creation of the dataset, by the innovation of this domain in Madagascar, and by the relevance of our result.