{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/visual-saliency-detection-based-on-multiscale","title":"Visual Saliency Detection Based on Multiscale Deep CNN Features","arxiv_id":"1609.02077","date":"2016-09-07","proceeding":null,"authors":["Guanbin Li","Yizhou Yu"],"abstract":"Visual saliency is a fundamental problem in both cognitive and computational\nsciences, including computer vision. In this paper, we discover that a\nhigh-quality visual saliency model can be learned from multiscale features\nextracted using deep convolutional neural networks (CNNs), which have had many\nsuccesses in visual recognition tasks. For learning such saliency models, we\nintroduce a neural network architecture, which has fully connected layers on\ntop of CNNs responsible for feature extraction at three different scales. The\npenultimate layer of our neural network has been confirmed to be a\ndiscriminative high-level feature vector for saliency detection, which we call\ndeep contrast feature. To generate a more robust feature, we integrate\nhandcrafted low-level features with our deep contrast feature. To promote\nfurther research and evaluation of visual saliency models, we also construct a\nnew large database of 4447 challenging images and their pixelwise saliency\nannotations. Experimental results demonstrate that our proposed method is\ncapable of achieving state-of-the-art performance on all public benchmarks,\nimproving the F- measure by 6.12% and 10.0% respectively on the DUT-OMRON\ndataset and our new dataset (HKU-IS), and lowering the mean absolute error by\n9% and 35.3% respectively on these two datasets.","url_abs":"http://arxiv.org/abs/1609.02077v1","url_pdf":"http://arxiv.org/pdf/1609.02077v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"visual-saliency-detection-based-on-multiscale","repo_url":"https://github.com/vlad-winter/FakePapers_NLP_Project","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"visual-saliency-detection-based-on-multiscale","repo_url":"https://github.com/vlad-winter/NLP20-FakePapers","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"saliency-detection","task_name":"Saliency Detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1609.02077","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}