Abstract: Existing occluded person re-identification (re-ID) methods mainly learn limited visual information for occluded pedestrians from images. However, textual information, which can describe ...

Abstract: Keyword spotting is often implemented by keyword classifier to the encoder in acoustic models, enabling the classification of predefined or open vocabulary keywords. Although keyword ...