图像字幕有时传达的信息超过其所能看到的图像（英文版）

英文标题：Image Captioners Sometimes Tell More Than Images They See中文摘要：本研究探讨了图像字幕生成器将原始图像信息转化为文本时信息的损失程度，并通过对比使用标准图像分类器和仅使用文本分类器进行的灾难图像分类任务的结果来评估了几种图像字幕生成模型，表明在某些情况下文本分类器可以比标准图像分类器实现更高的准确度，同时还表明将图像分类器与文本分类器进行融合可以提高准确性。英文摘要：Image captioning, a.k.a. "image-to-text," which generates descriptive textfrom given images, has been rapidly developing throughout the era of deeplearning. To what extent is the information in the original image preserved inthe descriptive text generated by an image captioner? To