Light-weighted Saliency Detection with Distinctively Lower Memory Cost and Model Size

TL;DR

This paper introduces a lightweight saliency detection method that significantly reduces memory and model size while maintaining competitive performance on benchmark datasets.

Contribution

The authors propose a novel, memory-efficient saliency detection approach with drastically lower runtime memory and model size compared to existing deep neural network methods.

Findings

Runtime memory cost is 42 to 99 times lower.

Model size is 63 to 129 times smaller.

Achieves competitive results on multiple benchmarks.

Abstract

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.