Light-weighted Saliency Detection with Distinctively Lower Memory Cost and Model Size
This work addresses the problem of excessive computational resources for saliency detection, which is important for applications requiring efficient low-level visual processing, but it is incremental as it builds on existing DNN-based methods.
The paper tackles the high computational cost of deep neural networks for saliency detection by proposing a light-weighted approach that significantly reduces runtime memory and model size, achieving competitive results on benchmark datasets with memory cost 42-99 times lower and model size 63-129 times smaller than previous methods.
Deep neural networks (DNNs) based saliency detection approaches have succeed in recent years, and improved the performance by a great margin via increasingly sophisticated network architecture. Despite the performance improvement, the computational cost is excessively high for such low level visual task. In this work, we propose a light-weighted saliency detection approach with distinctively lower runtime memory cost and model size. We evaluated the performance of our approach on multiple benchmark datasets, and achieved competitive results comparing with state-of-the-art methods on multiple metrics. We also evaluated the computational cost of our approach with multiple measurements. The runtime memory cost of our approach is 42 to 99 times fewer comparing with the previous DNNs based methods. The model size of our approach is 63 to 129 times smaller, and takes less than 1 Megabytes storage space with out any deep compression technique.