MRRC: multiple role representation crossover interpretation for image captioning with R-CNN feature distribution composition (FDC)

While image captioning through machines requires structured learning and basis for interpretation, improvement multiple context understanding processing in a meaningful way. This research provides novel concept combination impacts many applications to deal with visual features as an equivalence of descriptions objects, activities, events. There are three components our architecture: Feature Distribution Composition (FDC) Layer Attention, Multiple Role Representation Crossover (MRRC) Attention Layer, the Language Decoder. FDC helps generating weighted attention from RCNN features, MRRC acts intermediate representation next word attention. A Decoder estimation likelihood probable sentence. We demonstrated effectiveness FDC, MRRC, regional object feature attention, reinforcement effective generate better captions images. The performance model enhanced previous performances 35.3% created new standard theory generation based on logic, interpretability, contexts.