15篇「ICCV2021 Oral」最新論文搶先看！看當(dāng)下計算機視覺在研究什么？

來源：專知

本文約800字，建議閱讀10分鐘 
本文整理來自Twitter、arXiv、知乎放出來的15篇最新ICCV Oral論文，方便大家搶先閱覽！

最近計算機視覺三大頂會之一ICCV2021接收結(jié)果已經(jīng)公布，本次ICCV共計 6236 篇有效提交論文，其中有 1617 篇論文被接收，接收率為25.9%。小編在這里整理來自Twitter、arXiv、知乎放出來的15篇最新ICCV Oral論文，方便大家搶先閱覽！這些論文包括目標(biāo)檢測、域自適應(yīng)、因果推理、語義分割等。

1. BARF:束調(diào)整神經(jīng)輻射場，BARF: Bundle-Adjusting Neural Radiance Fields

論文鏈接：

https://chenhsuanlin.bitbucket.io/bundle-adjusting-NeRF/

2. 端到端多模態(tài)理解的調(diào)制檢測，MDETR : Modulated Detection for End-to-End Multi-Modal Understanding

論文鏈接：

https://www.zhuanzhi.ai/paper/945f0402f0332c41872bb7869e490be3

3. 稠密對應(yīng)無監(jiān)督學(xué)習(xí)，Warp Consistency for Unsupervised Learning of Dense Correspondences

論文鏈接：

https://www.zhuanzhi.ai/paper/678dd14c5c3ed4b6ab0c4c782ed0f135

4. 目標(biāo)檢測和實例分割的Rank & Sort損失，Rank & Sort Loss for Object Detection and Instance Segmentation

論文鏈接：

https://arxiv.org/abs/2002.12213

5. 遞歸條件高斯的有序無監(jiān)督域自適應(yīng), Recursively Conditional Gaussian for Ordinal Unsupervised Domain Adaptation

論文鏈接：

https://www.zhuanzhi.ai/paper/64e91d0014516f1556b0b8101808d141

6. SimROD:一種簡單的魯棒目標(biāo)檢測自適應(yīng)方法，SimROD: A Simple Adaptation Method for Robust Object Detection

論文鏈接：

https://www.zhuanzhi.ai/paper/5ddf35892e95179e384ff22f84e52821

7. 殘差對數(shù)似然估計的人體姿態(tài)回歸，Human Pose Regression with Residual Log-likelihood Estimation

論文鏈接：

https://www.zhuanzhi.ai/paper/e029b16a9f5bdfbf2b83993a1e4d3be2

8. 無監(jiān)督域適應(yīng)的運輸因果機制，Transporting Causal Mechanisms for Unsupervised Domain Adaptation

論文鏈接：

https://www.zhuanzhi.ai/paper/57f86ec4e25735fa6a744ec9d1747851

9. 深度假檢測的自一致性學(xué)習(xí)，Learning Self-Consistency for Deepfake Detection

論文鏈接：

https://www.zhuanzhi.ai/paper/5115974ee34bc53b0e76fce5b5f5b264

10. 自監(jiān)督對應(yīng)學(xué)習(xí)，Rethinking Self-supervised Correspondence Learning: A Video Frame-level Similarity Perspective

論文鏈接：

https://www.zhuanzhi.ai/paper/57095349f92f887dd0c496af3986197e

11. 研究語義分割中無監(jiān)督領(lǐng)域自適應(yīng)的魯棒性，Exploring Robustness of Unsupervised Domain Adaptation in Semantic Segmentation

論文鏈接：

https://www.zhuanzhi.ai/paper/0b6a434549580ea778cdf816522b5784

12. 用于穩(wěn)健姿態(tài)估計的三維人體運動模型，HuMoR: 3D Human Motion Model for Robust Pose Estimation

論文鏈接：

https://www.zhuanzhi.ai/paper/9f76eaa62f8a456a6acddb107e1c1569

13. 半監(jiān)督語義分割，Re-distributing Biased Pseudo Labels for Semi-supervised Semantic Segmentation: A Baseline Investigation

論文鏈接：

https://www.zhuanzhi.ai/paper/8400abaa3b0718f15fdf5b531c86eeda

14. 通過跨域集成的魯棒性，Robustness via Cross-Domain Ensembles

論文鏈接：

https://www.zhuanzhi.ai/paper/2ab2815739ff76f5e17ac41dbb537175

15. 弱監(jiān)督物體定位路由再思考，Just Ask: Learning to Answer Questions from Millions of Narrated Videos

論文鏈接：

https://www.zhuanzhi.ai/paper/fe5d9d7861cec92597a33dbf3178d776

編輯：黃繼彥