Salient Object Detection Dataset

Datasets Used in Our Research

This research utilizes the following five datasets for salient object detection:

1. PASCAL-S Dataset
2. • Description: PASCAL-S is a dataset designed for salient object detection. It consists of 850 images taken from the PASCAL VOC 2010 validation set, featuring multiple salient objects within various scenes.
   • Link: PASCAL-S Dataset
3. HKU-IS Dataset
4. • Description: The HKU-IS dataset, created by Guanbin Li and Yizhou Yu from The University of Hong Kong, is used for salient object detection. It has been used in the development of Deep Contrast Learning methods for improving saliency map accuracy. This dataset can be downloaded from Google Drive or Baidu Cloud.
   • Key Papers:
   • • Li, G., & Yu, Y. (2016). Deep Contrast Learning for Salient Object Detection. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, June 2016.
     • Li, G., & Yu, Y. (2016). Visual Saliency Detection Based on Multiscale Deep CNN Features, IEEE Transactions on Image Processing (TIP), 25(11), 5012-5024.
     • Li, G., & Yu, Y. (2015). Visual Saliency Based on Multiscale Deep Features, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, June 2015.
   • Link: HKU-IS Dataset
5. DUT-OMRON Dataset
6. • Description: The DUT-OMRON Image Dataset consists of high-quality images selected from a variety of natural scenes. The dataset is designed for research in both salient object detection and eye fixation prediction. It provides pixel-wise ground truth, bounding box ground truth, and eye-fixation ground truth for each image. The images in this dataset are more complex and challenging compared to other existing datasets.
   • Key Features:
   • • Images resized to 500x500 or 600x600 pixels.
     • Contains 1 or more salient objects with complex backgrounds.
     • Ground truth data was gathered from 5 participants with normal vision.
   • Link: DUT-OMRON Dataset
7. CCD Changsha, China Dataset
8. • Description: This dataset contains images of salient objects captured in Changsha, Hunan, China, using an iPhone 11 camera with a 12-megapixel resolution. The dataset includes images taken both indoors and outdoors, showcasing various objects under different lighting conditions. The goal is to assess the generalizability of the proposed salient object detection scheme.
   • Key Features:
   • • 622 images collected across the campus of Central South University and the city of Changsha.
     • Images include multiple salient objects and vary in lighting and resolution.
   • Link: CCD Changsha Dataset
9. Additional Salient Object Detection Datasets
10. • Description: This section references datasets that may have been used in supplementary work for training and testing saliency detection models. Detailed links to each dataset are provided in the research papers and supplementary materials. These datasets often provide specific focus on certain aspects, such as edge sharpness or object boundary detection.