I continued my research about word2vec. I wanted to have a tool that convert physical object into semantics — via Word2vec in order to compare them. It would then be a simple intuitive way to made comparaison between terms. So this system is based with two camera one add words one minus words of thing it see. The detection system is based on YOLO (which has great performance) and use the database COCO (with around 80 categories of detectable object). An Interface (made in html/css) shows the operation in real time and does transition whenever there is a changing equation.

the interface