Tsinghua University

Zhaoqun Cui

Peng Wang

Rongbo Zhang

The present application provides a multi-modality based building image retrieval method, device and apparatus. The method includes the steps of: acquiring a building example image set, a positive prompt word set, and a negative prompt word set inputted by a user; extracting features from the building example image set, the positive prompt word set and the negative prompt word set, and calculating a mean image feature, a mean positive feature, and a mean negative feature; and calculating correlations between the building example image set and various building images in a building image library according to the mean image feature, the mean positive feature, and the mean negative feature. In the present application, image retrieval is performed using data of text and image modalities, and the obtained target images have a high correlation with building example images, requiring no manual screening, and further enhancing work efficiency.

Growing community of inventors

Beijing, China

Zhaoqun Cui