This paper introduces a novel approach for salient object detection by incorporating short connections into the skip-layer structures within the Holistically-Nested Edge Detector (HED) architecture. The proposed method leverages multi-level and multi-scale features extracted from Fully Convolutional Neural Networks (FCNs) to enhance the performance of saliency detection. By introducing short connections, the framework combines high-level semantic information from deeper layers with rich low-level spatial information from shallower layers, resulting in more accurate and efficient saliency maps. The method achieves state-of-the-art results on five widely tested benchmarks, demonstrating superior efficiency (0.08 seconds per image), effectiveness, and simplicity compared to existing algorithms. The paper also conducts a comprehensive analysis on the impact of training data on performance, providing a more unified and fair training set for future research.This paper introduces a novel approach for salient object detection by incorporating short connections into the skip-layer structures within the Holistically-Nested Edge Detector (HED) architecture. The proposed method leverages multi-level and multi-scale features extracted from Fully Convolutional Neural Networks (FCNs) to enhance the performance of saliency detection. By introducing short connections, the framework combines high-level semantic information from deeper layers with rich low-level spatial information from shallower layers, resulting in more accurate and efficient saliency maps. The method achieves state-of-the-art results on five widely tested benchmarks, demonstrating superior efficiency (0.08 seconds per image), effectiveness, and simplicity compared to existing algorithms. The paper also conducts a comprehensive analysis on the impact of training data on performance, providing a more unified and fair training set for future research.