… image identification task. This suggests that our system could be used as-is to generate Web
image captions … (which are the source of Conceptual Captions annotations). Qualitatively, …

… Since there is no benchmark dataset for evaluating cross-domain image captioning
systems… of image captioning, image synthesis, multi-modal learning, cross-domain sequence …

… where plenty of labeled training data (ie image-text pairs) are readily available … image
captioning methods to a new domain. In this paper, we propose a cross-domain image captioning …

… that images and captions … cross-domain image captioning. We define style as a syntactic
structure rather than users’preferences like “positive”, “negative” etc. in stylized image captioning…

… novel cross-domain image captioning approach that couples an adaptive image captioning
… the adaptation of the image captioning model by generating pseudo image-sentence pairs in …

… image captioning models, we propose a new benchmark called Domain Generalization for
Image Captioning (… Multitask learning for crossdomain image captioning. IEEE TMM, 2018. 2 […

… To achieve the goal of crossdomain image captioning, we propose a novel method consisting
of two main components. The first component is a standard CNN-RNN-based captioner (…

… the cross-domain image captioning, our training goal becomes the following. For the image I
… In order to realize the goal of cross-domain image captioning, we propose a simple method …

… Then, we introduce Cross Domain Stylized (CDS) loss to perform cross-domain style
alignment. The proposed CDS loss combines style reward information from each separate style …

… In section 2, we review related works related to research in the cross-domain image
captioning and instance-based transfer learning. In section 3, we introduce the proposed model …