主催: 電気・情報関係学会九州支部連合大会委員会
共催: 佐賀大学
会議名: 2021年度電気・情報関係学会九州支部連合大会
回次: 74
開催地: オンライン開催（大会本部：佐賀大学本庄キャンパス）
開催日: 2021/09/24 - 2021/09/25
Similar images may share close features such as color and shape. But the similarity could also be built on entirely different images. We plan to fully utilize textual information for image similarity evaluation. The first textual information is scene text, which is available by using the CTC dataset. The second one is the object label, which is available by the COCO dataset. It is possible to have a feature vector by converting the textual information into semantic vectors by word2vec. Next, we observe the property of the feature vectors via clustering analysis. Similarly, we extract features from images and implement clustering analysis.
