Learned image embedding with a more semantic text embedding to achieve theĭesired knowledge transfer from seen to unseen classes. Inspired by recent advances in image-textįoundation models (e.g., CLIP) on zero-shot scenarios, we explicitly align the Learn new abstract concepts of the sketch domain and improve cross-domain
Specifically, we insert simple and lightweight domain adapters to We present an effective ``Adapt and Align'' approach to address the keyĬhallenges. However, these efforts are inadequate in adaptingĭomains and transferring knowledge from seen to unseen classes. Models with various side information and learning strategies to learn a compactįeature space that is shared between the sketch and photo domains and bridges Download a PDF of the paper titled Adapt and Align to Improve Zero-Shot Sketch-Based Image Retrieval, by Shiyin Dong and 4 other authors Download PDF Abstract: Zero-shot sketch-based image retrieval (ZS-SBIR) is challenging due to theĬross-domain nature of sketches and photos, as well as the semantic gap between
0 kommentar(er)
