Channel and spatial attention mechanism for fashion image captioning

<div class="page" title="Page 1"><div class="layoutArea"><div class="column"><p><span lang="EN-US">Image captioning aims to automatically generate one or more description sentences for a given input image. Most of the existing methods use encoder-decoder model which mainly focus on recognizing and capturing relationship between objects appearing in However, when generating captions fashion images, it is important not only describe items their relationships, but also mention attribute features clothes (shape, texture, style, fabric, more). In this study, novel proposed image task can capture relationship, features. Two different attention mechanisms (spatial-attention channel-wise attention) incorporated traditional model, dynamically interprets caption sentence multi-layer feature map addition depth dimension map. We evaluate our architecture Fashion-Gen using three metrics (CIDEr, ROUGE-L, BLEU-1), achieve scores 89.7, 50.6 45.6, respectively. Based experiments, method shows significant performance improvement fashion-image captioning, outperforms other state-of-the-art methods.</span></p></div></div></div>