[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_344":345},"routes/_layout",{"_11":254},"routes/entity.$type.$id",{"_11":12,"_13":14,"_15":16,"_17":18,"_19":20,"_21":22,"_23":24,"_25":26,"_27":28},"dehydratedState",{"_60":61,"_62":63},"type","person","id","person_A5100727358","section","articles","entity",{"_15":16,"_29":30,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":38,"_39":-5,"_40":41,"_42":43,"_44":45,"_46":47,"_48":-5,"_49":-5,"_50":-5,"_51":-5,"_52":-5,"_53":-5,"_54":-5,"_55":-5,"_56":-5,"_57":-5,"_58":30,"_59":-5},"canonical","https://www.hugmap.com/entity/person/person_A5100727358/articles","siteBaseUrl","https://www.hugmap.com","sectionCount",0,"jsonLd","{\"@context\":\"https://schema.org\",\"@graph\":[{\"name\":\"Yunhe Wang\",\"url\":\"https://www.hugmap.com/entity/person/person_A5100727358/articles\",\"@type\":\"Person\",\"jobTitle\":null},{\"@type\":\"CollectionPage\",\"name\":\"Yunhe Wang 相关文章\",\"description\":\"Yunhe Wang 在 HugMap 中的相关文章，含结构化关联信息与详情链接。\",\"url\":\"https://www.hugmap.com/entity/person/person_A5100727358/articles\",\"inLanguage\":\"zh-CN\"},{\"@type\":\"BreadcrumbList\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"首页\",\"item\":\"https://www.hugmap.com/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"人物\",\"item\":\"https://www.hugmap.com/search?type=person\"},{\"@type\":\"ListItem\",\"position\":3,\"name\":\"Yunhe Wang\",\"item\":\"https://www.hugmap.com/entity/person/person_A5100727358\"},{\"@type\":\"ListItem\",\"position\":4,\"name\":\"相关文章\",\"item\":\"https://www.hugmap.com/entity/person/person_A5100727358/articles\"}]}]}","name","Yunhe Wang","nameZh","aliases","description","descriptionZh","status","active","externalIds","{\"openalex\":\"A5100727358\",\"orcid\":\"0000-0002-2709-4946\"}","properties","source","openalex","createdAt","2026-05-14T01:44:04.266799607","updatedAt","2026-05-23T08:26:08.239980356","orcid","0000-0002-2709-4946","title","affiliation","researchArea","paperCount","citationCount","homepage","linkedin","twitter","github","hindex","displayName","displayDescription","mutations",[],"queries",[64],{"_65":66,"_67":68,"_69":70,"_71":72},"dehydratedAt",1781142563728,"state",{"_74":75,"_76":77,"_78":66,"_79":-5,"_80":26,"_81":26,"_82":26,"_83":-5,"_84":-5,"_85":86,"_35":87,"_88":89},"queryKey",[73,19,14,16],"queryHash","[\"portal\",\"entity\",\"person\",\"person_A5100727358\"]","portal","data",{"_19":20,"_90":91,"_92":93,"_94":95,"_96":97,"_98":99,"_100":101,"_18":102},"dataUpdateCount",1,"dataUpdatedAt","error","errorUpdateCount","errorUpdatedAt","fetchFailureCount","fetchFailureReason","fetchMeta","isInvalidated",false,"success","fetchStatus","idle","entityType","Person","taxonomies",[],"papers",[158,159,160,161,162,163],"companies",[103,104,105],"events",[],"patents",[],[],{"_15":142,"_29":143,"_31":-5,"_32":144,"_33":-5,"_34":-5,"_35":36,"_37":145,"_39":-5,"_40":41,"_42":146,"_44":147,"_114":115,"_116":-5,"_117":148,"_119":-5,"_121":-5,"_122":-5,"_123":-5,"_124":-5,"_125":149,"_127":-5,"_128":-5,"_129":-5,"_58":143,"_59":-5},{"_15":134,"_29":135,"_31":-5,"_32":136,"_33":-5,"_34":-5,"_35":36,"_37":137,"_39":-5,"_40":41,"_42":138,"_44":139,"_114":140,"_116":-5,"_117":-5,"_119":120,"_121":-5,"_122":-5,"_123":-5,"_124":-5,"_125":-5,"_127":-5,"_128":-5,"_129":-5,"_58":135,"_59":-5},{"_15":106,"_29":107,"_31":-5,"_32":108,"_33":109,"_34":-5,"_35":36,"_37":110,"_39":-5,"_40":111,"_42":112,"_44":113,"_114":115,"_116":-5,"_117":118,"_119":120,"_121":-5,"_122":-5,"_123":-5,"_124":-5,"_125":126,"_127":-5,"_128":-5,"_129":-5,"_58":107,"_59":109},"company_I4210159102","Huawei Technologies (China)",[130,131,132,107,133],"Huawei Technologies Co., Ltd. is a Chinese multinational corporation and technology company headquartered in Longgang, Shenzhen, Guangdong. Its main product lines include telecommunications equipment, consumer electronics, electric vehicle autonomous driving systems, and rooftop solar power products. The company was founded in Shenzhen in 1987 by Ren Zhengfei, a veteran officer of the People's Liberation Army (PLA).","{\"ror\":\"00cmhce21\",\"openalex\":\"I2250955327\",\"wikidata\":\"https://www.wikidata.org/wiki/Q160120\"}","openalex,graphrag","2026-05-14T01:44:04.247034696","2026-06-01T06:45:23.816734106","companyType","COMPANY","foundedDate","headquarters","Düsseldorf, Germany","country","CN","employeeCount","fundingStage","totalFunding","valuation","website","http://www.huawei.com/de/","stockSymbol","uscc","crunchbaseUrl","Huawei Technologies (Germany)","Huawei Poland","Huawei Technologies (Poland)","华为技术有限公司","company_I99065089","Tsinghua University",[135,141],"{\"ror\":\"03cve4549\",\"openalex\":\"I99065089\"}","2026-05-14T00:28:38.40459851","2026-06-01T04:14:47.734032813","UNIVERSITY","清华大学","company_I4390039265","PRG S&Tech (South Korea)",[150,151,143,152,153,154,155,156,157],"{\"ror\":\"02sr2ee22\",\"openalex\":\"I4390039265\"}","2026-05-14T05:09:33.474510545","2026-06-01T12:39:33.367831775","Busan, South Korea","http://prgst.com","PRG S & Tech Inc.","PRG S&Tech","PRG S&Tech Inc.","PRG Science & Technology Inc.","PRG Science and Technology","PRG Science and Technology Inc.","Pioneer of Rare Genopathies via Science and Technology","피알지에스앤텍",{"_15":241,"_29":242,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":243,"_39":-5,"_40":41,"_42":244,"_44":245,"_169":246,"_171":-5,"_172":247,"_48":242,"_174":248,"_176":249,"_178":250,"_180":-5,"_52":251,"_183":-5,"_184":252,"_186":-5,"_187":86,"_58":242,"_59":-5},{"_15":228,"_29":229,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":230,"_39":-5,"_40":41,"_42":231,"_44":232,"_169":233,"_171":-5,"_172":234,"_48":229,"_174":235,"_176":199,"_178":200,"_180":-5,"_52":236,"_183":-5,"_184":237,"_186":-5,"_187":86,"_58":229,"_59":-5},{"_15":218,"_29":219,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":220,"_39":-5,"_40":41,"_42":221,"_44":222,"_169":223,"_171":-5,"_172":224,"_48":219,"_174":225,"_176":177,"_178":179,"_180":181,"_52":226,"_183":-5,"_184":227,"_186":-5,"_187":86,"_58":219,"_59":-5},{"_15":204,"_29":205,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":206,"_39":-5,"_40":41,"_42":207,"_44":208,"_169":209,"_171":-5,"_172":210,"_48":205,"_174":211,"_176":212,"_178":213,"_180":-5,"_52":214,"_183":-5,"_184":215,"_186":-5,"_187":86,"_58":205,"_59":-5},{"_15":191,"_29":192,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":193,"_39":-5,"_40":41,"_42":194,"_44":195,"_169":196,"_171":-5,"_172":197,"_48":192,"_174":198,"_176":199,"_178":200,"_180":-5,"_52":201,"_183":-5,"_184":202,"_186":-5,"_187":86,"_58":192,"_59":-5},{"_15":164,"_29":165,"_31":-5,"_32":-5,"_33":-5,"_34":-5,"_35":36,"_37":166,"_39":-5,"_40":41,"_42":167,"_44":168,"_169":170,"_171":-5,"_172":173,"_48":165,"_174":175,"_176":177,"_178":179,"_180":181,"_52":182,"_183":-5,"_184":185,"_186":-5,"_187":86,"_58":165,"_59":-5},"paper_W4313142416","Multimodal Token Fusion for Vision Transformers","{\"openalex\":\"W4313142416\",\"doi\":\"10.1109/cvpr52688.2022.01187\"}","2026-05-23T08:26:06.189536027","2026-05-23T08:26:06.189538687","doi","10.1109/cvpr52688.2022.01187","arxivId","openalexId","W4313142416","abstractText","Many adaptations of transformers have emerged to address the single-modal vision tasks, where self-attention modules are stacked to handle input sources like images. Intuitively, feeding multiple modalities of data to vision transformers could improve the performance, yet the innermodal attentive weights may be diluted, which could thus greatly undermine the final performance. In this paper, we propose a multimodal token fusion method (TokenFusion), tailored for transformer-based vision tasks. To effectively fuse multiple modalities, TokenFusion dynamically detects uninformative tokens and substitute these tokens with projected and aggregated inter-modal features. Residual positional alignment is also adopted to enable explicit utilization of the inter-modal alignments after fusion. The design of TokenFusion allows the transformer to learn correlations among multimodal features, while the single-modal transformer architecture remains largely intact. Extensive experiments are conducted on a variety of homogeneous and heterogeneous modalities and demonstrate that TokenFusion surpasses state-of-the-art methods in three typical vision tasks: multimodal image-to-image translation, RGB-depth semantic segmentation, and 3D object detection with point cloud and images. Code will be released ¹ ¹ https://github.com/huawei-noah/noah-research ² ² https://gitee.com/mindspore/models/tree/master/research/cv/TokenFusion.","publishedDate","2022-06-01","publishedYear",2022,"venue","2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)",220,"authors","keywords",[188,189,190],"pdfUrl","topVenue","Advanced Neural Network Applications","Robotics and Sensor-Based Localization","Visual Attention and Saliency Detection","paper_W3176459575","Distilling Object Detectors via Decoupled Features","{\"openalex\":\"W3176459575\",\"doi\":\"10.1109/cvpr46437.2021.00219\"}","2026-05-22T15:17:20.302958156","2026-05-22T15:17:20.302960629","10.1109/cvpr46437.2021.00219","W3176459575","Knowledge distillation is a widely used paradigm for inheriting information from a complicated teacher network to a compact student network and maintaining the strong performance. Different from image classification, object detectors are much more sophisticated with multiple loss functions in which features that semantic information rely on are tangled. In this paper, we point out that the information of features derived from regions excluding objects are also essential for distilling the student detector, which is usually ignored in existing approaches. In addition, we elucidate that features from different regions should be assigned with different importance during distillation. To this end, we present a novel distillation algorithm via decoupled features (DeFeat) for learning a better student detector. Specifically, two levels of decoupled features will be processed for embedding useful information into the student, i.e., decoupled features from neck and decoupled proposals from classification head. Extensive experiments on various detectors with different backbones show that the proposed DeFeat is able to surpass the state-of-the-art distillation methods for object detection. For example, DeFeat improves ResNet50 based Faster R-CNN from 37.4% to 40.9% mAP, and improves ResNet50 based RetinaNet from 36.5% to 39.7% mAP on COCO benchmark. Code will be released ^1,2 .","2021-06-01",2021,244,[188,190,203],"Advanced Image and Video Retrieval Techniques","paper_W2982802130","Data-Free Learning of Student Networks","{\"openalex\":\"W2982802130\",\"doi\":\"10.1109/iccv.2019.00361\"}","2026-05-20T00:57:24.535732848","2026-05-20T00:57:24.536116367","10.1109/iccv.2019.00361","W2982802130","Learning portable neural networks is very essential for computer vision for the purpose that pre-trained heavy deep models can be well applied on edge devices such as mobile phones and micro sensors. Most existing deep neural network compression and speed-up methods are very effective for training compact deep models, when we can directly access the training dataset. However, training data for the given deep network are often unavailable due to some practice problems (\\eg privacy, legal issue, and transmission), and the architecture of the given network are also unknown except some interfaces. To this end, we propose a novel framework for training efficient deep neural networks by exploiting generative adversarial networks (GANs). To be specific, the pre-trained teacher networks are regarded as a fixed discriminator and the generator is utilized for derivating training samples which can obtain the maximum response on the discriminator. Then, an efficient network with smaller model size and computational complexity is trained using the generated data and the teacher network, simultaneously. Efficient student networks learned using the proposed Data-Free Learning (DFL) method achieve 92.22% and 74.47% accuracies without any training data on the CIFAR-10 and CIFAR-100 datasets, respectively. Meanwhile, our student network obtains an 80.56% accuracy on the CelebA benchmark.","2019-10-01",2019,366,[188,216,217],"Domain Adaptation and Few-Shot Learning","Video Surveillance and Tracking Methods","paper_W4312950730","CMT: Convolutional Neural Networks Meet Vision Transformers","{\"openalex\":\"W4312950730\",\"doi\":\"10.1109/cvpr52688.2022.01186\"}","2026-05-14T20:24:30.392434944","2026-05-20T20:38:57.199573341","10.1109/cvpr52688.2022.01186","W4312950730","Vision transformers have been successfully applied to image recognition tasks due to their ability to capture long-range dependencies within an image. However, there are still gaps in both performance and computational cost between transformers and existing convolutional neural networks (CNNs). In this paper, we aim to address this issue and develop a network that can outperform not only the canonical transformers, but also the high-performance convolutional models. We propose a new transformer based hybrid network by taking advantage of transformers to capture long-range dependencies, and of CNNs to extract local information. Furthermore, we scale it to obtain a family of models, called CMTs, obtaining much better trade-off for accuracy and efficiency than previous CNN-based and transformer-based models. In particular, our CMT-S achieves 83.5% top-1 accuracy on ImageNet, while being 14x and 2x smaller on FLOPs than the existing DeiT and EfficientNet, respectively. The proposed CMT-S also generalizes well on CIFAR10 (99.2%), CIFAR100 (91.7%), Flowers (98.7%), and other challenging vision datasets such as COCO (44.3% mAP), with considerably less computational cost.",840,[188,216,203],"paper_W3171125843","Pre-Trained Image Processing Transformer","{\"openalex\":\"W3171125843\",\"doi\":\"10.1109/cvpr46437.2021.01212\"}","2026-05-14T04:40:37.393964977","2026-05-21T06:58:11.192504529","10.1109/cvpr46437.2021.01212","W3171125843","As the computing power of modern hardware is increasing strongly, pre-trained deep learning models (e.g., BERT, GPT-3) learned on large-scale datasets have shown their effectiveness over conventional methods. The big progress is mainly contributed to the representation ability of transformer and its variant architectures. In this paper, we study the low-level computer vision task (e.g., denoising, super-resolution and deraining) and develop a new pre-trained model, namely, image processing transformer (IPT). To maximally excavate the capability of transformer, we present to utilize the well-known ImageNet benchmark for generating a large amount of corrupted image pairs. The IPT model is trained on these images with multi-heads and multi-tails. In addition, the contrastive learning is introduced for well adapting to different image processing tasks. The pre-trained model can therefore efficiently employed on desired task after fine-tuning. With only one pre-trained model, IPT outperforms the current state-of-the-art methods on various low-level benchmarks. Code is available at https://github.com/huawei-noah/Pretrained-IPT and https://gitee.com/mindspore/mindspore/tree/master/model_zoo/research/cv/IPT",1975,[238,239,240],"Advanced Image Processing Techniques","Image and Signal Denoising Methods","Image Enhancement Techniques","paper_W3035414587","GhostNet: More Features From Cheap Operations","{\"openalex\":\"W3035414587\",\"doi\":\"10.1109/cvpr42600.2020.00165\"}","2026-05-14T01:44:04.208128465","2026-05-21T03:28:02.207528821","10.1109/cvpr42600.2020.00165","W3035414587","Deploying convolutional neural networks (CNNs) on embedded devices is difficult due to the limited memory and computation resources. The redundancy in feature maps is an important characteristic of those successful CNNs, but has rarely been investigated in neural architecture design. This paper proposes a novel Ghost module to generate more feature maps from cheap operations. Based on a set of intrinsic feature maps, we apply a series of linear transformations with cheap cost to generate many ghost feature maps that could fully reveal information underlying intrinsic features. The proposed Ghost module can be taken as a plug-and-play component to upgrade existing convolutional neural networks. Ghost bottlenecks are designed to stack Ghost modules, and then the lightweight GhostNet can be easily established. Experiments conducted on benchmarks demonstrate that the proposed Ghost module is an impressive alternative of convolution layers in baseline models, and our GhostNet can achieve higher recognition performance (e.g. 75.7% top-1 accuracy) than MobileNetV3 with similar computational cost on the ImageNet ILSVRC-2012 classification dataset. Code is available at https://github.com/huawei-noah/ghostnet.","2020-06-01",2020,4320,[188,216,253],"COVID-19 diagnosis using AI",{"_60":255,"_62":256},[],[257],{"_65":258,"_67":259,"_69":260,"_71":261},1781142562158,{"_74":263,"_76":77,"_78":258,"_79":-5,"_80":26,"_81":26,"_82":26,"_83":-5,"_84":-5,"_85":86,"_35":87,"_88":89},[73,262],"[\"portal\",\"industries\"]","industries",[264,265,266,267,268,269,270,271,272,273,274,275],{"_15":337,"_29":338,"_31":-5,"_32":339,"_33":340,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":341,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":338,"_59":340},{"_15":333,"_29":334,"_31":-5,"_32":-5,"_33":335,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":336,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":334,"_59":335},{"_15":327,"_29":328,"_31":-5,"_32":329,"_33":330,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":331,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":328,"_59":330},{"_15":323,"_29":324,"_31":-5,"_32":-5,"_33":325,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":326,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":324,"_59":325},{"_15":319,"_29":320,"_31":-5,"_32":-5,"_33":321,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":322,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":320,"_59":321},{"_15":313,"_29":314,"_31":-5,"_32":315,"_33":316,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":317,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":314,"_59":316},{"_15":308,"_29":309,"_31":-5,"_32":-5,"_33":310,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":311,"_288":26,"_289":312,"_290":26,"_291":-5,"_58":309,"_59":310},{"_15":304,"_29":305,"_31":-5,"_32":-5,"_33":306,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":307,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":305,"_59":306},{"_15":300,"_29":301,"_31":-5,"_32":-5,"_33":302,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":303,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":301,"_59":302},{"_15":296,"_29":297,"_31":-5,"_32":-5,"_33":298,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":299,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":297,"_59":298},{"_15":292,"_29":293,"_31":-5,"_32":-5,"_33":294,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":295,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":293,"_59":294},{"_15":276,"_29":277,"_31":-5,"_32":-5,"_33":278,"_34":-5,"_35":36,"_37":-5,"_39":-5,"_40":279,"_42":-5,"_44":-5,"_280":281,"_282":-5,"_283":77,"_284":285,"_286":287,"_288":26,"_289":86,"_290":26,"_291":-5,"_58":277,"_59":278},"tax_ind_smartphone","智能手机","Smartphone","manual","taxonomyType","INDUSTRY","code","level","parentId","tax_ind_root","path","科技行业/智能手机","sortOrder","isLeaf","entityCount","children","tax_ind_robot","机器人","Robotics","科技行业/机器人","tax_ind_wearable","智能穿戴","Smart Wearables","科技行业/智能穿戴","tax_ind_energy","新能源与智能制造","New Energy & Smart Manufacturing","科技行业/新能源与智能制造","tax_ind_biotech","生物科技","Biotech","科技行业/生物科技","tax_ind_quantum","量子计算","Quantum Computing","科技行业/量子计算",true,"tax_ind_blockchain","区块链",[318],"Blockchain","科技行业/区块链","Web3","tax_ind_enterprise","企业软件","Enterprise Software","科技行业/企业软件","tax_ind_security","网络安全","Cybersecurity","科技行业/网络安全","tax_ind_semi","半导体",[332],"Semiconductor","科技行业/半导体","芯片","tax_ind_cloud","云计算","Cloud Computing","科技行业/云计算","tax_ind_ai","人工智能",[342,343],"Artificial Intelligence","科技行业/人工智能","AI","AI 技术","env",{"_346":24},"SITE_BASE_URL"]

Yunhe Wang 人物

Yunhe Wang · 相关文章