3D【9】人臉重建:Joint3D Face Reconstruction and Dense Alignment with Position Map Regression
該論文同時預測人臉特徵點和3D mesh的頂點座標x，y，z。人臉特徵點沒什麼好說的，現在已經有很多成功的案例了。主要關注該論文是如何預測頂點座標的。其實大概想想，就能夠大致確定直接預測頂點座標是有不少困難的。首先是，你不能把所有的頂點拉成一個向量，用全連接層預測；要用卷積層作爲輸出的話，mesh又不能直接表示成一個固定大小的張量（這正好是用cnn來預測所需要解決的問題，比如將mesh立方體化）。
相關文章
- 1. 3D人臉重建：《Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network》
- 2. 人臉對齊（十八）--Joint Face Alignment and 3D Face Reconstruction
- 3. 3D【11】人臉重建：Disentangling Features in 3D Face Shapes for Joint Face Reconstruction and Recognition
- 4. 人臉對齊--Face Alignment by Explicit Shape Regression
- 5. 三維人臉重建論文收集
- 6. 人臉對齊--One Millisecond Face Alignment with an Ensemble of Regression Trees
- 7. (Face Alignment) One Millisecond Face Alignment with an Ensemble of Regression Trees 筆記
- 8. 微軟三維人臉重建論文總結——《Accurate 3D Face Reconstruction with Weakly-Supervised Learning》
- 9. PRNet：人臉3D重建與密集對齊
- 10. End-to-end 3D face reconstruction with deep neural networks
- 更多相關文章...
- • Scala Map(映射) - Scala教程
- • XSL-FO table-and-caption 對象 - XSL-FO 教程
- • RxJava操作符（七）Conditional and Boolean
- • 再有人問你分佈式事務，把這篇扔給他
相關標籤/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
歡迎關注本站公眾號,獲取更多信息
相關文章
- 1. 3D人臉重建：《Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network》
- 2. 人臉對齊（十八）--Joint Face Alignment and 3D Face Reconstruction
- 3. 3D【11】人臉重建：Disentangling Features in 3D Face Shapes for Joint Face Reconstruction and Recognition
- 4. 人臉對齊--Face Alignment by Explicit Shape Regression
- 5. 三維人臉重建論文收集
- 6. 人臉對齊--One Millisecond Face Alignment with an Ensemble of Regression Trees
- 7. (Face Alignment) One Millisecond Face Alignment with an Ensemble of Regression Trees 筆記
- 8. 微軟三維人臉重建論文總結——《Accurate 3D Face Reconstruction with Weakly-Supervised Learning》
- 9. PRNet：人臉3D重建與密集對齊
- 10. End-to-end 3D face reconstruction with deep neural networks