{"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/stc-a-simple-to-complex-framework-for-weakly","title":"STC: A Simple to Complex Framework for Weakly-supervised Semantic Segmentation","arxiv_id":"1509.03150","date":"2015-09-10","proceeding":null,"authors":["Yunchao Wei","Xiaodan Liang","Yunpeng Chen","Xiaohui Shen","Ming-Ming Cheng","Jiashi Feng","Yao Zhao","Shuicheng Yan"],"abstract":"Recently, significant improvement has been made on semantic object\nsegmentation due to the development of deep convolutional neural networks\n(DCNNs). Training such a DCNN usually relies on a large number of images with\npixel-level segmentation masks, and annotating these images is very costly in\nterms of both finance and human effort. In this paper, we propose a simple to\ncomplex (STC) framework in which only image-level annotations are utilized to\nlearn DCNNs for semantic segmentation. Specifically, we first train an initial\nsegmentation network called Initial-DCNN with the saliency maps of simple\nimages (i.e., those with a single category of major object(s) and clean\nbackground). These saliency maps can be automatically obtained by existing\nbottom-up salient object detection techniques, where no supervision information\nis needed. Then, a better network called Enhanced-DCNN is learned with\nsupervision from the predicted segmentation masks of simple images based on the\nInitial-DCNN as well as the image-level annotations. Finally, more pixel-level\nsegmentation masks of complex images (two or more categories of objects with\ncluttered background), which are inferred by using Enhanced-DCNN and\nimage-level annotations, are utilized as the supervision information to learn\nthe Powerful-DCNN for semantic segmentation. Our method utilizes $40$K simple\nimages from Flickr.com and 10K complex images from PASCAL VOC for step-wisely\nboosting the segmentation network. Extensive experimental results on PASCAL VOC\n2012 segmentation benchmark well demonstrate the superiority of the proposed\nSTC framework compared with other state-of-the-arts.","url_abs":"http://arxiv.org/abs/1509.03150v2","url_pdf":"http://arxiv.org/pdf/1509.03150v2.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":"stc-a-simple-to-complex-framework-for-weakly","repo_url":"https://github.com/shimoda-uec/ssdd","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"salient-object-detection","task_name":"RGB Salient Object Detection"},{"task_slug":"salient-object-detection-1","task_name":"Salient Object Detection"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"weakly-supervised-semantic-segmentation-1","task_name":"Weakly supervised Semantic Segmentation"},{"task_slug":"weakly-supervised-semantic-segmentation","task_name":"Weakly-Supervised Semantic Segmentation"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1509.03150","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}