Semantic Understanding for Intelligent Computational Photography
Selection and composition are at the core of the image editing and computational photography. Adding the high-level understanding of visual media can facilitate intelligent automatic selection and composition. This can be viewed as an artificial intelligence for image and video editing. In this talk, I introduce my recent efforts on stepping stones between scene context understanding and media editing, such as video summaries via text description (WACV18), automatic cinemagraphs (ICCV17), sound-based scene region selection (CVPR18), semantic soft segmentation (SIGGRAPH18).