ObjectPatchNet: Towards scalable and semantic image annotation and retrieval

Open Webpage

Shiliang Zhang, Qi Tian, Gang Hua, Qingming Huang, Wen Gao

Published: 2014, Last Modified: 13 Nov 2024Comput. Vis. Image Underst. 2014EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•We build ObjectPatchNet (OPN) from large-scale loosely annotated Internet images.•OPN preserves regionlevel semantic labels and contextual clues for image annotation.•OPN could be utilized as visual words with semantics for scalable image retrieval.•Though not exhaustively tuned, OPN performs decently in these two applications.•OPN could be an open platform for different applications and further improvements.