W-Net Plus: Dnn For Spatial Saliency Prediction In Videos

Deep neural networks (DNNs) have been employed recently to identify saliency in images and videos. Different models have been suggested to predict spatiotemporal as well as temporal and spatial saliency. In this study, a model called W-net Plus is proposed for predicting spatial saliency in videos. The newly proposed model not only translates information from the encoder to the decoder but also from the decoder to the encoder, and it is inspired by U -net architecture. The proposed model has been evaluated by the author(s) on various data sets, and they have obtained noteworthy findings for the F - Max, structural similarity (S-measure), and Mean Absolute Error (MAE) measurement parameters, demonstrating its effectiveness and survivability. No stage of the training involves the utilization of transfer learning in the work.