{"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/dynamic-structured-semantic-propagation","title":"Dynamic-structured Semantic Propagation Network","arxiv_id":"1803.06067","date":"2018-03-16","proceeding":"CVPR 2018 6","authors":["Xiaodan Liang","Hongfei Zhou","Eric Xing"],"abstract":"Semantic concept hierarchy is still under-explored for semantic segmentation\ndue to the inefficiency and complicated optimization of incorporating\nstructural inference into dense prediction. This lack of modeling semantic\ncorrelations also makes prior works must tune highly-specified models for each\ntask due to the label discrepancy across datasets. It severely limits the\ngeneralization capability of segmentation models for open set concept\nvocabulary and annotation utilization. In this paper, we propose a\nDynamic-Structured Semantic Propagation Network (DSSPN) that builds a semantic\nneuron graph by explicitly incorporating the semantic concept hierarchy into\nnetwork construction. Each neuron represents the instantiated module for\nrecognizing a specific type of entity such as a super-class (e.g. food) or a\nspecific concept (e.g. pizza). During training, DSSPN performs the\ndynamic-structured neuron computation graph by only activating a sub-graph of\nneurons for each image in a principled way. A dense semantic-enhanced neural\nblock is proposed to propagate the learned knowledge of all ancestor neurons\ninto each fine-grained child neuron for feature evolving. Another merit of such\nsemantic explainable structure is the ability of learning a unified model\nconcurrently on diverse datasets by selectively activating different neuron\nsub-graphs for each annotation at each step. Extensive experiments on four\npublic semantic segmentation datasets (i.e. ADE20K, COCO-Stuff, Cityscape and\nMapillary) demonstrate the superiority of our DSSPN over state-of-the-art\nsegmentation models. Moreoever, we demonstrate a universal segmentation model\nthat is jointly trained on diverse datasets can surpass the performance of the\ncommon fine-tuning scheme for exploiting multiple domain knowledge.","url_abs":"http://arxiv.org/abs/1803.06067v1","url_pdf":"http://arxiv.org/pdf/1803.06067v1.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":[],"tasks":[{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"universal-segmentation","task_name":"Universal Segmentation"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"bottleneck-residual-block","method_name":"Bottleneck Residual Block"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semantic-segmentation-on-ade20k","task":"Semantic Segmentation","dataset":"ADE20K","model":"DSSPN (ResNet-101)","rank_in_archive_order":207,"of":235,"metrics":{"Validation mIoU":"43.68"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-ade20k-val","task":"Semantic Segmentation","dataset":"ADE20K val","model":"DSSPN (ResNet-101)","rank_in_archive_order":89,"of":95,"metrics":{"mIoU":"43.68"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-cityscapes","task":"Semantic Segmentation","dataset":"Cityscapes test","model":"DSSPN (ResNet-101)","rank_in_archive_order":62,"of":105,"metrics":{"Mean IoU (class)":"77.8%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.06067","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}