Deep Visual Semantic Embedding with Text Data Augmentation and Word Embedding Initialization

Figure 2

Performance for percent of the dataset used for training.

(a)

(b)