| MSRA | Learning to Detect a Salient Object, CVPR 2007. | Not Found | Not Specific | Image set A: 20840 images, Image set B: 5000 images selected from A with less ambiguity. (Training set: Randomly select 2000 from set A, and 1000 from set B) |
| MSRA-B | Salient Object Detection: A Discriminative Regional Feature Integration Approach, CVPR 2013. | Homepage | Image set B of MSRA | 2500 images for training, 500 for validation, and 2000 for testing (full list can be found at the bottom of the project page.) |
| MSRA10K<br>(THUS10000) | Global Contrast based Salient Region Detection, PAMI 2015. | Homepage | MSRA | -- |
| SED1/2 | Image Segmentation by Probabilistic Bottom-up Aggregation and Cue Integration, CVPR 2007. | Homepage | Unknown | 200 images, 1 or 2 clear objects. |
| MSRA-1000 | Frequency-tuned Salient Region Detection, CVPR 2009. | Homepage | MSRA | Contour label. Work that be evaluated on this dataset has less power of persuasion, suggested by CMM. |
| CSSD | Hierarchical Saliency Detection, CVPR 2013. | Homepage | BSD300, VOC dataset and internet. | Complex Scene Saliency Dataset(CSSD). 200 images.(Normally used for test.) |
| ECSSD | Hierarchical Image Saliency Detection on Extended CSSD, PAMI 2016. | Homepage | Internet | Semantically meaningful but structurally complex images. 1000 images.(Normally used for test.) |
| THUR | SalientShape: Group Saliency in Image Collections, The Visual Computing 2014. | Homepage | Flickr | 6000+ images for 5 categories: butterfly, coffe mug, dog jump, griaffe, and plane. |
| PASCAL-S | The Secrets of Salient Object Detection, CVPR 2014. | Homepage | the validation set of the PASCAL VOC 2010 | Relative saliency between salient instances. 850 images(Normally used for test.). |
| HKU-IS | Deep Contrast Learning for Salient Object Detection, CVPR 2015. | Homepage | Unknown | 2500 images for training, 500 for validation, and 1447 for testing. |
| Judd-DB | What is a salient object? A dataset and a baseline model for salient object detection, TIP 2015 | Unknown | Judd | 900 images. |
| DUT-OMRON | Saliency Detection Via Graph-Based Manifold Ranking, CVPR 2013. | Homepage | SUN database | 5019 images for testing. More details could be found in their FCV paper. |
| DUTS | Learning to Detect Salient Objects with Image-level Supervision, CVPR 2017. | Homepage | ImageNet DET datasets | 10553 training images and 5019 test images. |
| ILSO | Instance-Level Salient Object Segmentation, CVPR 2017. | Homepage | ECSSD, DUT-OMRON, HKU-IS, and MSO datasets. | Two-thirds of the chosen images contain multiple occluded salient object instances while the remaining one-third consists of images with no salient regions, a single salient object instance or multiple salient instances without occlusion. Totally 1000 images, 500 for training, 200 for validation, and 300 for testing. |
| XPIE | What is and what is not a salient object? learning salient object detector by ensembling linear exemplar regressors, CVPR 2017. | Homepage | Panoramio, ImageNet, and two fixation datasets. | 10000 images. |
| SOC | Salient Objects in Clutter: Bringing Salient Object Detection to the Foreground, ECCV 2018. | Homepage | MSCOCO | 6K images(3.6K for training, 1.2K for validation, 1.2K for testing). Accurate and clear boundary label. instance-level segmentation. Visual attributes for result evaluation. |
| SOS | Salient Object Subitizing, CVPR 2015. | Homepage | 6900images in total. 2064 from the COCO, 2141 from the VOC07, 1610 from the ImageNet, 1085 from the SUN. | This dataset is built for predicting the existence and the number of salient objects in a scene. MSO is a subset of SOS, which is used for testing purpose. |
| BSDS | A Database of Human Segmented Natural Images and its Application to Evaluating Segmentation Algorithms and Measuring Ecological Statistics, ICCV 2001. | Homepage | Unknown | Original version: 200 training images, 100 testing images. Extented version: 200 test images were added. |