Simple vs complex temporal recurrences for video saliency prediction. In British Machine Vision Conference (BMVC). Cardiff, Wales / UK.: British Machine Vision Association; 2019. (1.79 MB) .
Abstract
This paper investigates modifying an existing neural network architecture for static saliency prediction using two types of recurrences that integrate information from the temporal domain. The first modification is the addition of a ConvLSTM within the architecture, while the second is a computationally simple exponential moving average of an internal convolutional state. We use weights pre-trained on the SALICON dataset and fine-tune our model on DHF1K. Our results show that both modifications achieve state-of-the-art results and produce similar saliency maps.
- BMVC 2019 (acceptance rate=28%)
- Project site and Source code
- Paper on arXiv.
- Blog post by Akis Linardos.