{"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/semi-supervised-semantic-segmentation-using-2","title":"Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels","arxiv_id":"2203.03884","date":"2022-03-08","proceeding":"CVPR 2022 1","authors":["Yuchao Wang","Haochen Wang","Yujun Shen","Jingjing Fei","Wei Li","Guoqiang Jin","Liwei Wu","Rui Zhao","Xinyi Le"],"abstract":"The crux of semi-supervised semantic segmentation is to assign adequate pseudo-labels to the pixels of unlabeled images. A common practice is to select the highly confident predictions as the pseudo ground-truth, but it leads to a problem that most pixels may be left unused due to their unreliability. We argue that every pixel matters to the model training, even its prediction is ambiguous. Intuitively, an unreliable prediction may get confused among the top classes (i.e., those with the highest probabilities), however, it should be confident about the pixel not belonging to the remaining classes. Hence, such a pixel can be convincingly treated as a negative sample to those most unlikely categories. Based on this insight, we develop an effective pipeline to make sufficient use of unlabeled data. Concretely, we separate reliable and unreliable pixels via the entropy of predictions, push each unreliable pixel to a category-wise queue that consists of negative samples, and manage to train the model with all candidate pixels. Considering the training evolution, where the prediction becomes more and more accurate, we adaptively adjust the threshold for the reliable-unreliable partition. Experimental results on various benchmarks and training settings demonstrate the superiority of our approach over the state-of-the-art alternatives.","url_abs":"https://arxiv.org/abs/2203.03884v2","url_pdf":"https://arxiv.org/pdf/2203.03884v2.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":"semi-supervised-semantic-segmentation-using-2","repo_url":"https://github.com/Haochen-Wang409/U2PL","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"semi-supervised-semantic-segmentation","task_name":"Semi-Supervised Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-2","task":"Semi-Supervised Semantic Segmentation","dataset":"Cityscapes 12.5% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, AEL)","rank_in_archive_order":16,"of":33,"metrics":{"Validation mIoU":"76.48%"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-1","task":"Semi-Supervised Semantic Segmentation","dataset":"Cityscapes 25% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, AEL)","rank_in_archive_order":12,"of":30,"metrics":{"Validation mIoU":"78.51%"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-8","task":"Semi-Supervised Semantic Segmentation","dataset":"Cityscapes 50% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, AEL)","rank_in_archive_order":14,"of":23,"metrics":{"Validation mIoU":"79.12%"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-22","task":"Semi-Supervised Semantic Segmentation","dataset":"Cityscapes 6.25% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, AEL)","rank_in_archive_order":13,"of":18,"metrics":{"Validation mIoU":"74.90%"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-10","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 1464 labels","model":"U2PL (DeepLab v3+ with ResNet-101)","rank_in_archive_order":12,"of":17,"metrics":{"Validation mIoU":"79.5"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-28","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 183 labeled","model":"U2PL (DeepLab v3+ with ResNet-101)","rank_in_archive_order":14,"of":16,"metrics":{"Validation mIoU":"69.2"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-9","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 25% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, CutMix)","rank_in_archive_order":10,"of":27,"metrics":{"Validation mIoU":"79.3"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-29","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 366 labeled","model":"U2PL (DeepLab v3+ with ResNet-101)","rank_in_archive_order":13,"of":15,"metrics":{"Validation mIoU":"73.7"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-15","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 50%","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, CutMix)","rank_in_archive_order":3,"of":14,"metrics":{"Validation mIoU":"80.5%"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-30","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 732 labeled","model":"U2PL (DeepLab v3+ with ResNet-101)","rank_in_archive_order":14,"of":16,"metrics":{"Validation mIoU":"76.2"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-27","task":"Semi-Supervised Semantic Segmentation","dataset":"PASCAL VOC 2012 92 labeled","model":"U2PL (DeepLab v3+ with ResNet-101)","rank_in_archive_order":15,"of":17,"metrics":{"Validation mIoU":"68.0"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-4","task":"Semi-Supervised Semantic Segmentation","dataset":"Pascal VOC 2012 12.5% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, CutMix)","rank_in_archive_order":10,"of":38,"metrics":{"Validation mIoU":"79.01%"},"uses_additional_data":false},{"leaderboard":"/sota/semi-supervised-semantic-segmentation-on-21","task":"Semi-Supervised Semantic Segmentation","dataset":"Pascal VOC 2012 6.25% labeled","model":"U2PL (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K, CutMix)","rank_in_archive_order":10,"of":19,"metrics":{"Validation mIoU":"77.21"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2203.03884","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2203.03884"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Haochen-Wang409/U2PL","reach":null}],"summary":{"ran_fixture":1},"by_repo_kind":{"official":{"samples":1,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"00f3224921ae6a19","entry":"compute_unsupervised_loss","repo":"Haochen-Wang409/U2PL","repo_kind":"official","path":"u2pl/utils/loss_helper.py","file_url":"https://github.com/Haochen-Wang409/U2PL/blob/HEAD/u2pl/utils/loss_helper.py","link_basis":"first_harvest_node","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"00f3224921ae6a19"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}