整理:AI算法与图像处理
CVPR2022论文和代码整理:https://github.com/DWCTOD/CVPR2022-Papers-with-Code-Demo
ECCV2022论文和代码整理:https://github.com/DWCTOD/ECCV2022-Papers-with-Code-Demo
最新成果demo展示:
标题:Parallel Inversion of Neural Radiance Fields for Robust Pose Estimation
论文:https://pnerfp.github.io/
提出了一种基于快速神经辐射场(NeRF)的并行优化方法,用于估计6-DoF目标姿态。给定目标的单个观察到的RGB图像,我们可以通过最小化从fastNeRF模型渲染的像素与观察到的图像中的像素之间的残差来预测相机的平移和旋转。我们将基于动量的相机外部优化程序集成到Instant Neural Graphics Primitives中,这是最近一个非常快速的NeRF实现。通过在姿态估计任务中引入并行蒙特卡罗采样,我们的方法克服了局部极小值,并在更广泛的搜索空间中提高了效率。我们还说明了采用更稳健的基于像素的损失函数来减少误差的重要性。实验证明,我们的方法可以在合成和真实基准上实现更好的泛化和鲁棒性。
最新论文整理
ECCV2022
Updated on : 24 Oct 2022
total number : 6
PoseScript: 3D Human Poses from Natural Language
- 论文/Paper: http://arxiv.org/pdf/2210.11795
- 代码/Code: None
Distilling the Undistillable: Learning from a Nasty Teacher
- 论文/Paper: http://arxiv.org/pdf/2210.11728
- 代码/Code: https://github.com/surgan12/nastyattacks
Context-Enhanced Stereo Transformer
- 论文/Paper: http://arxiv.org/pdf/2210.11719
- 代码/Code: None
Photo-realistic 360 Head Avatars in the Wild
- 论文/Paper: http://arxiv.org/pdf/2210.11594
- 代码/Code: None
Rethinking Learning Approaches for Long-Term Action Anticipation
- 论文/Paper: http://arxiv.org/pdf/2210.11566
- 代码/Code: https://github.com/nmegha2601/anticipatr
Overexposure Mask Fusion: Generalizable Reverse ISP Multi-Step Refinement
- 论文/Paper: http://arxiv.org/pdf/2210.11511
- 代码/Code: None
CVPR2022
NeurIPS
Updated on : 24 Oct 2022
total number : 2
Diffusion Visual Counterfactual Explanations
- 论文/Paper: http://arxiv.org/pdf/2210.11841
- 代码/Code: https://github.com/valentyn1boreiko/dvces
XC: Exploring Quantitative Use Cases for Explanations in 3D Object Detection
- 论文/Paper: http://arxiv.org/pdf/2210.11590
- 代码/Code: None