CVPR论文总结
Best Paper:
A Theory of Fermat Paths for Non-Line-of-Sight Shape Reconstruction
非视线形状重建的费马路径理论 搞机器人的那一套不是很懂,但感觉到机器人很火,后续发展方向
卡内基梅隆 辛书冕
Best Student:
Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation
面向语言视觉导航(VLN)的强化交叉模型匹配和半监督模仿学习,做视觉语言导航的,也不是很懂,技术交叉有点多,也可以算是机器人方向
杜克大学 王鑫
经典论文奖 李飞飞~ ImageNet 大名鼎鼎 名副其实
最佳论文荣誉提名
英伟达的 A Style-Based Generator Architecture for Generative Adversarial Networks
谷歌的 Learning the Depths of Moving People by Watching Frozen People
人脸篇:
7. MVF-Net: Multi-View 3D Face Morphable Model Regression
8. Towards High-Fidelity Nonlinear 3D Face Morphable Model
9. GANFIT: Generative Adversarial Network Fitting for High Fidelity 3D Face Reconstruction
10. AdaptiveFace: Adaptive Margin and Sampling for Face Recognition
通过设置可以学习的Margin来解决数据样本不均导致不同类别的类内方差异所带来的问题
GAN篇:
1. Geometry-Consistent Generative Adversarial Networks for One-Sided Unsupervised Domain Mapping oral
2. Label-Noise Robust Generative Adversarial Networks oral
3. Sphere Generative Adversarial Network Based on Geometric Moment Matching oral
4. A Style-Based Generator Architecture for Generative Adversarial Networks oral
5. Constrained Generative Adversarial Networks for Interactive Image Generation oral
1.Unsupervised Face Normalization with Extreme Pose and Expression in the Wild.
UFN 目的为了得到矫正后的人脸,包括pose,expression,lighting的矫正。矫正后的人脸再去做分类性能得到明显提升。
NAS篇:
MnasNet: Platform-Aware Neural Architecture Search for Mobile
表情篇:
A Compact Embedding for Facial Expression Similarity
Facial Emotion Distribution Learning by Exploiting Low-Rank Label Correlations Locally