20篇「ICCV2021 Oral」最新论文抢先看！看当下计算机视觉在研究什么？

最近计算机视觉三大顶会之一ICCV2021接收结果已经公布，本次ICCV共计 6236 篇有效提交论文，其中有 1617 篇论文被接收，接收率为25.9%。专知在这里整理来自Twitter、arXiv、知乎放出来的20篇最新ICCV Oral论文，方便大家抢先阅览！这些论文包括目标检测、域自适应、因果推理、语义分割等。

1. BARF:束调整神经辐射场，BARF: Bundle-Adjusting Neural Radiance Fields

https://www.zhuanzhi.ai/paper/b63a7cac32419f3d5cd9601cc0f41cff https://chenhsuanlin.bitbucket.io/bundle-adjusting-NeRF/

2. 端到端多模态理解的调制检测，MDETR : Modulated Detection for End-to-End Multi-Modal Understanding

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

3. 稠密对应无监督学习，Warp Consistency for Unsupervised Learning of Dense Correspondences

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

4. 目标检测和实例分割的Rank & Sort损失，Rank & Sort Loss for Object Detection and Instance Segmentation

https://arxiv.org/abs/2002.12213

5. 递归条件高斯的有序无监督域自适应, Recursively Conditional Gaussian for Ordinal Unsupervised Domain Adaptation

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

6. SimROD:一种简单的鲁棒目标检测自适应方法，SimROD: A Simple Adaptation Method for Robust Object Detection

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

7. 残差对数似然估计的人体姿态回归，Human Pose Regression with Residual Log-likelihood Estimation

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

8. 无监督域适应的运输因果机制，Transporting Causal Mechanisms for Unsupervised Domain Adaptation

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

9. 深度假检测的自一致性学习，Learning Self-Consistency for Deepfake Detection

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

10. 自监督对应学习，Rethinking Self-supervised Correspondence Learning: A Video Frame-level Similarity Perspective

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

11. 研究语义分割中无监督领域自适应的鲁棒性，Exploring Robustness of Unsupervised Domain Adaptation in Semantic Segmentation

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

12. 用于稳健姿态估计的三维人体运动模型，HuMoR: 3D Human Motion Model for Robust Pose Estimation

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

13. 半监督语义分割，Re-distributing Biased Pseudo Labels for Semi-supervised Semantic Segmentation: A Baseline Investigation

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

14. 通过跨域集成的鲁棒性，Robustness via Cross-Domain Ensembles https://www.zhuanzhi.ai/paper/2ab2815739ff76f5e17ac41dbb537175

15. 弱监督物体定位路由再思考，Just Ask: Learning to Answer Questions from Millions of Narrated Videos

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