sual feed-forward attention mechanisms as 'bottom-up'. Most conventional visual attention mechanisms used in image captioning and VQA are of the top-down ...
確定! 回上一頁