Frame-level advert localisation

With the increase in multimedia videos, the applications of videos and images to leverage targeted advertisements are becoming significantly diverse in reaching specific audiences. Targeted advertisements are achieved by detecting and replacing an existing advertisement within an image frame with a new advert. In this work, we benchmarked our proposed model with other billboard video datasets to ensure that the proposed model is not specific to the dataset initially used. We evaluated our method by comparing it to other semantic segmentation algorithms using a publicly available dataset of outdoor scenes annotated with a binary billboard map. The results suggest that AdSegNet consistently shows the highest Average IoUf across all videos, ranging from 0.864 to 0.913, indicating superior segmentation performance. Our findings demonstrate that by benchmarking our proposed model with a new dataset, we gain insights into its generalization ability and confirm its performance across diverse datasets. This underscores its reliability and broad applicability beyond the specific dataset on which it was initially trained.