Identifying Candidate Spaces for Advert Implantation

摘要: Virtual advertising is an important and promising feature in the area of online advertising. It involves integrating adverts onto live or recorded videos for product placements targeted advertisements. Such integration primarily done by video editors post-production stage, which cumbersome time-consuming. Therefore, it to automatically identify candidate spaces a frame, wherein new can be implanted. The space should match scene perspective, also have high quality experience according human subjective judgment. In this paper, we propose use bespoke neural net that assist identifying spaces. We benchmark our approach against several deep-learning architectures on large-scale image dataset outdoor scenes. Our work first its kind multimedia augmented reality applications, achieves best results.