Nano1337/CLOVE-scores-english-VQAnotCLIP

该数据集包含100个样本，主要用于图像和文本描述的相关性分析。每个样本包含用户ID（uid）、图像（image）、文本描述（caption）、CLIP评分（CLIPscore）、VQA评分（VQAscore）等字段。CLIP评分和VQA评分分别表示图像与文本描述之间的相关性，以及图像与问题回答之间的相关性。数据集还提供了这些评分的百分位数（CLIPscore_percentile、VQAscore_percentile）以及百分位数差异（percentile_difference）。数据集仅包含一个训练集（train），大小为3582349.0字节。This dataset contains 100 samples and is primarily used for analyzing the correlation between images and text descriptions. Each sample includes fields such as user ID (uid), image, caption, CLIPscore, VQAscore, etc. The CLIPscore and VQAscore represent the correlation between the image and text description, and the correlation between the image and question answering, respectively. The dataset also provides percentiles for these scores (CLIPscore_percentile, VQAscore_percentile) and the percentile difference (percentile_difference). The dataset contains only a training set (train) with a size of 3582349.0 bytes.