Dual-level Collaborative Transformer for Image Captioning

Descriptive region features extracted by object detection networks have played an important role in the recent advancements of image captioning. However, they are still criticized for lack contextual information and fine-grained details, which contrast merits traditional grid features. In this paper, we introduce a novel Dual-Level Collaborative Transformer (DLCT) network to realize complementary advantages two Concretely, DLCT, these first processed Dual-way Self Attenion (DWSA) mine their intrinsic properties, where Comprehensive Relation Attention component is also introduced embed geometric information. addition, propose Locality-Constrained Cross module address semantic noises caused direct fusion features, alignment graph constructed accurately align reinforce To validate our model, conduct extensive experiments on highly competitive MS-COCO dataset, achieve new state-of-the-art performance both local online test sets, i.e., 133.8% CIDEr Karpathy split 135.4% official split.