Simple vs complex temporal recurrences for video saliency prediction
arXiv:1907.01869
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 conceptually 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. Source code is available at https://git.io/fjPiB.
Accepted at BMVC 2019
References in corpus (8)
- Adam: A Method for Stochastic Optimization
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
- Improving neural networks by preventing co-adaptation of feature detectors
- Two-Stream Convolutional Networks for Action Recognition in Videos
- SalGAN: Visual Saliency Prediction with Generative Adversarial Networks
- Shallow and Deep Convolutional Networks for Saliency Prediction
- Temporal Activity Detection in Untrimmed Videos with Recurrent Neural Networks
Cited by in corpus (6)
- Motion-Attentive Transition for Zero-Shot Video Object Segmentation
- Temporal-Spatial Feature Pyramid for Video Saliency Detection
- Attention-SLAM: A Visual Monocular SLAM Learning from Human Gaze
- Video Saliency Prediction Using Enhanced Spatiotemporal Alignment Network
- Toward Better Understanding of Saliency Prediction in Augmented 360 Degree Videos
- MediaEval 2019: Concealed FGSM Perturbations for Privacy Preservation