可迁移表示 | ViLab https://vilab.team/tag/%E5%8F%AF%E8%BF%81%E7%A7%BB%E8%A1%A8%E7%A4%BA/ 可迁移表示 Hugo Blox Builder (https://hugoblox.com)en-usFri, 01 Jan 2021 00:00:00 +0000 https://vilab.team/media/icon_hu2896232876136423579.png 可迁移表示 https://vilab.team/tag/%E5%8F%AF%E8%BF%81%E7%A7%BB%E8%A1%A8%E7%A4%BA/ Task-independent knowledge makes for transferable representations for generalized zero-shot learning https://vilab.team/publication/task-independent-knowledge-makes-for-transferable-representa/ Fri, 01 Jan 2021 00:00:00 +0000 https://vilab.team/publication/task-independent-knowledge-makes-for-transferable-representa/ <p>本文针对广义零样本学习（GZSL）中可见类与未见类之间的表示偏差问题，提出利用任务无关知识来学习可迁移的视觉表示。通过在大规模辅助数据上预训练或引入外部知识，使模型捕获与类别标签无关的通用特征，从而提升对未见类别的识别能力。在多个基准数据集上验证了方法的有效性。</p>