A wide range of image captioning models has been developed, achieving significant improvement based on popular metrics, such as BLEU, CIDEr, and SPICE. However, although the generated captions can accurately describe the image, they are generic for similar images and lack distinctiveness, i.e., cannot properly describe the uniqueness of each image. In this paper, we aim to improve the distinctiveness of image captions through training with sets of similar images. First, we propose a distinctiveness metric -- between-set CIDEr (CIDErBtw) to evaluate the distinctiveness of a caption with respect to those of similar images. Our metric shows that the human annotations of each image are not equivalent based on distinctiveness. Thus we propose several new training strategies to encourage the distinctiveness of the generated caption for each image, which are based on using CIDErBtw in a weighted loss function or as a reinforcement learning reward. Finally, extensive experiments are conducted, showing that our proposed approach significantly improves both distinctiveness (as measured by CIDErBtw and retrieval metrics) and accuracy (e.g., as measured by CIDEr) for a wide variety of image captioning baselines. These results are further confirmed through a user study.

中文翻译：

---

比较和重新加权：使用相似图像集的独特图像字幕

已经开发了广泛的图像字幕模型，基于流行的指标（例如 BLEU、CIDEr 和 SPICE）实现了显着改进。然而，虽然生成的字幕可以准确地描述图像，但它们对于相似图像是通用的，缺乏独特性，即不能正确描述每幅图像的唯一性。在本文中，我们的目标是通过训练相似图像集来提高图像标题的独特性。首先，我们提出了一个独特性度量——集间 CIDEr (CIDErBtw) 来评估标题相对于相似图像的独特性。我们的度量标准表明，基于独特性，每张图像的人工注释并不等效。因此，我们提出了几种新的训练策略，以鼓励每张图像生成的标题的独特性，它们基于在加权损失函数中使用 CIDErBtw 或作为强化学习奖励。最后，进行了大量实验，表明我们提出的方法显着提高了各种图像字幕基线的独特性（由 CIDErBtw 和检索指标衡量）和准确性（例如，由 CIDEr 衡量）。这些结果通过用户研究得到进一步证实。