We have conducted research on massive multimedia datasets including text, 3D shape models, and images, and extract valuable pieces of information. Main focus of our research has been on feature extraction, information retrieval, clustering, classification, segmentation, and automatic annotation or tagging to multimedia.The following picture shows the outlook of our research activities.
Research on Multimedia Retrieval, Classification, Segmentation, and Automatic Annotation
In this research, we use three-dimensional shape models, images, videos, and annotated texts, in order to apply semantic search, “ambiguous” search, partial-matching search, and classification. The basic idea is to extract salient features that represent an object robust and invariant under translation, rotation, scaling, and other similar operations. We boast of the world-top-level search performance for 3D shape models. In 2013, we were the world number one in the KINECT track of SHREC2013, where SHREC denotes Shape Retrieval Contest.
↑ Sample picture of similarity search of 3D objects
Web Mining, Data Mining, and Text Mining Research
Web is a resource including a billion of stones and a handful of gems. Web mining is an emerging research field, attempting to find “gems” on the Web. Out research on Web mining includes Web content mining, Web spam detection, Web link analysis, and Web usability monitoring. In addition we are conducting research on blog/microblog analysis, selected news extraction, personalized news delivery, opinion mining, sentiment analysis, information recommendation and filtering.