The codes for transferring knowledge for image captioning task based on PyTorch. Some of the codes are forked from the work of RuotianLuo.
The attributes are firstly defined and extracted for each image in the source domain and the corresponding features are used in the target domain in order to generate correct sentences.