We propose CLIP-Lite, an information efficient method for visual representation learning by feature alignment with textual annotations. Compared ...
確定! 回上一頁