Transformer拿下CV顶会大奖,微软亚研获ICCV 2021最佳论文
共 6619字,需浏览 14分钟
·
2021-10-19 01:05
点击“凹凸域”,马上关注
更多内容、请置顶或星标
ICCV 2021 全部奖项已经公布,来自微软亚洲研究院的研究者获得 ICCV 2021 马尔奖(最佳论文);最佳学生论文奖由苏黎世联邦理工学院、微软研究者获得。
![](https://filescdn.proginn.com/d5d6e29d4f5ed789b4d815d9898b6397/1df74c0ed7d048814dc129c7d20da999.webp)
![](https://filescdn.proginn.com/661f0c65686856e621867580e73b78c5/0cef2674fa851a6ab0f2ec67c4955384.webp)
![](https://filescdn.proginn.com/6100b547536c25708ca81ea5b77f9756/d556a770787871c7a93e87d2fac51cef.webp)
![](https://filescdn.proginn.com/2c9d7004bb24e56c846f9bd4ee409539/446e382c16e72ca1cab4e8bbad9bc2b3.webp)
获奖论文:Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
作者机构:微软亚洲研究院
论文地址:https://arxiv.org/pdf/2103.14030.pdf
项目地址:https://github.com/microsoft/Swin-Transformer
![](https://filescdn.proginn.com/2da4fe968cf90721b2100f1412045039/63ff9b59eb14c732d1db7bb087cbbecc.webp)
![](https://filescdn.proginn.com/89526b9dbe5d3d85b946bf72057dec61/8169b49390de58cb688c8558b8312c75.webp)
![](https://filescdn.proginn.com/d0ddc055a5fd6ff98a35480b49e33699/060897c2092679576cadba83ea37c440.webp)
![](https://filescdn.proginn.com/26a6d45ece80cc960a0ad788e84441ba/8f9d6a397b442f93b9db98af2aa459ef.webp)
获奖论文:Pixel-Perfect Structure-from-Motion with Featuremetric Refinement
作者机构:苏黎世联邦理工学院、微软
论文地址:https://arxiv.org/pdf/2108.08291.pdf
项目地址:github.com/cvg/pixel-perfect-sfm (http://github.com/cvg/pixel-perfect-sfm)
![](https://filescdn.proginn.com/19e15bc4a118913774069b27244f1c6b/8cd4cdd812f82d1005bff58dc02698a4.webp)
![](https://filescdn.proginn.com/7c17b9cdacf29c4f7fbca0568a75aec9/7e19d858b3e126ecfaaec4becc92360f.webp)
论文 1:Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields
作者机构:谷歌、加州大学伯克利分校
论文地址:https://arxiv.org/pdf/2103.13415.pdf
![](https://filescdn.proginn.com/14f29a9c73570bc7f46651754ebba21b/77020471ad5a6fa918c2ee7aa9a1f95e.webp)
论文 2:OpenGAN: Open-Set Recognition via Open Data Generation
作者机构:卡内基梅隆大学
论文地址:https://arxiv.org/pdf/2104.02939.pdf
![](https://filescdn.proginn.com/6e07fd1c1d9111e232dee3282479c133/8b7be7bf55668de3d5ccc187fbf34c78.webp)
论文 3:Viewing Graph Solvability via Cycle Consistency
作者机构:特伦托大学等
论文地址:https://openaccess.thecvf.com/content/ICCV2021/papers/Arrigoni_Viewing_Graph_Solvability_via_Cycle_Consistency_ICCV_2021_paper.pdf
![](https://filescdn.proginn.com/99989e3e04dfbcb9bd55dc2f85e9626b/f51aec27127d6fc2e2b6c96c08e24d40.webp)
论文 4:Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction
作者机构:Facebook AI 研究院、伦敦大学学院
论文地址:https://arxiv.org/pdf/2109.00512.pdf
![](https://filescdn.proginn.com/68d0d213f2703870514c1d47e9c73f95/5a0349f618fab1c1c270ee3647cc2b54.webp)
![](https://filescdn.proginn.com/4d5cbec27ba88c1a8d52b589dbd50f85/cee9d7dd74a326c4d5da505ff833e937.webp)
论文 1:《ORB: An efficient alternative to SIFT or SURF》
论文链接:https://ieeexplore.ieee.org/document/6126544
论文 2:《HMDB: A large video database for human motion recognition》
论文链接:https://ieeexplore.ieee.org/document/6126543
论文 3:《DTAM: Dense tracking and mapping in real-time》
论文链接:https://ieeexplore.ieee.org/document/6126513
![](https://filescdn.proginn.com/e991951155e7b03f1f27f7e8122bfa3e/644dba5761c8f078f4e2c0e52f677e55.webp)
![](https://filescdn.proginn.com/4eb0e05735852e595d1fdb28f658de1b/df597667cf2eea98ce79e6e42ceb1f39.webp)
![](https://filescdn.proginn.com/6df1f9afa4888ac038ebe3700d623795/ff18cb04f3e04d6d9bc1e81e3e1db557.webp)
![](https://filescdn.proginn.com/05bff3fc7da15978af6cc16a0f85705c/2ef0e54df00c828a608f4c1458416067.webp)
— END —
想要了解更多AI资讯
点这里👇关注我,记得标星呀~
请点击上方卡片,专注计算机人工智能方向的研究