BriefGPT.xyz
Feb, 2021
训练视觉Transformer进行图像检索
Training Vision Transformers for Image Retrieval
HTML
PDF
Alaaeldin El-Nouby, Natalia Neverova, Ivan Laptev, Hervé Jégou
TL;DR
本文提出一种基于变换器的图像检索方法，通过采用视觉变换器生成图像描述符并使用度量学习目标进行训练，结合对比损失和微分熵正则化，相比于卷积方法，提高了图像检索性能，特别是对于短向量表示和低分辨率图像。
Abstract
transformers
have shown outstanding results for natural language understanding and, more recently, for image classification. We here extend this work and propose a transformer-based approach for
image retrieval
:
→