Applying Deep Learning based Approach for Visual Saliency based Prediction

Abstract

The human eye and its vision is one of the most significant aspects of one's ability to understand the nature. Acquisition of visual intelligence and predicting visible attention is present terms used to describe how fast the human eye focuses on salient aspects in a scene while analysing it. In recent years, neural networks used to predict visual acuity. A deep neural network architecture based on transfer learning approach is proposed in this paper for visual saliency prediction. Proposed model extracts visual information from raw photos using convolutional layers in order predicting apparent stiffness. Additionally, the proposed model uses a VGG-16 architecture with semantic separation, using a pixel based separation layer for predicting the label of each pixel category in the inserted image. The proposed method is used in most cases to demonstrate its effectiveness.