Auto Image Comment via Deep Attention

In this paper, we propose an automatic image comment algorithm the state-of-the-art algorithm of image description. We employ an encoder-decoder model to extract the image feature in a deep learning style. In addition, an attention mechanism is added into the deep model for more efficient feature extraction. The model is trained and parameterized with the image comment dataset proposed in this paper. After training, the obtained deep model can generate comments automatically for a given image. In the experiment part, we evaluate the proposed method using both subjective way and objective way. We found that the proposed method can generate reasonable comments from the perspective of human beings. Meanwhile, the model was adjusted to achieve the best effect based on evaluation indexes such as BLEU and CIDEr.