最新论文汇总：深度补全

清疚 2022-12-22 06:26 134阅读 0赞

本文来自博主：流浪机器人，仅用于学术分享，侵删

原文地址：https://blog.csdn.net/qq\_26623879/article/details/108847218

# Self-supervised Sparse-to-Dense: Self-supervised Depth Completion from LiDAR and Monocular Camera #

![90c3cafc13eea26f3a992f4562e463e3.png][]

输入RGB以及稀疏LiDARs 由Depth 网络预测深度，然后在有深度的情况下用OPENCV 的 cv2.xfeatures2d.SIFT\_create()检测特征点，然后用cv2.solvePnPRansac解得Poset,通过Photometric Loss进行自监督

![f0e99968998b59bd1c654f695a935a1f.png][]

# Depth Estimation via Affinity Learned with Convolutional Spatial Propagation Network #

![f29b449b711af6d463a4da8066ea1acc.png][]

本文提出卷积稀疏传播网络，让周边像素向有深度的像素靠近。

# Deep Depth Completion of a Single RGB-D Image #

![0ef56f3865e6ca42ddd792a76db515e4.png][]

用RGB来估计遮挡边界和法线，然后用深度gt来优化整体深度。

![88d9fbf8809f42b027bcf45af44441e0.png][]

# Parse Geometry from a Line: Monocular Depth Estimation with Partial Laser Observation #

![17731877ef2d44ed40b29870b20a1b79.png][]

输入2D 雷达和单目图像，2D雷达生成的3D扫描平面，然后RGB+生成的参考深度来生成深度。

![58d00d8a4f3ec310cd9cfdf2d9991788.png][]

# DeepLiDAR: Deep Surface Normal Guided Depth Prediction for Outdoor Scene from Sparse LiDAR Data and Single Color Image #

##  ##

![14b024475aa3cb9b9db553b19862780e.png][]

本文是一篇通过稀疏深度，用网络估计法线，然后融合稀疏深度得到稠密深度，和单目深度估计深度融合，最终得到准确稠密深度。

![059b8aaf9cd1f9fafdc0b2b9956b8dfe.png][]

# Sparse and noisy LiDAR completion with RGB guidance and uncertainty #

![3169e8c8bcf8ff30e8a8220358d1c89f.png][]

# Unsupervised Depth Completion from Visual Inertial Odometry #

![3131011a97766c607a6224df1c99c37d.png][]

本文用VIO的稀疏深度作为监督，使用Depth estimation 网络估计深度

## 参考 ##

https://paperswithcode.com/task/depth-completion

[90c3cafc13eea26f3a992f4562e463e3.png]: /images/20221120/5cd7dc10bdc24d20b8e03e12c99bbc00.png
[f0e99968998b59bd1c654f695a935a1f.png]: /images/20221120/e4eea1bd0da640389750f68ee1631988.png
[f29b449b711af6d463a4da8066ea1acc.png]: /images/20221120/a02ea10c7e104ed0a6a1eadcd0d324c1.png
[0ef56f3865e6ca42ddd792a76db515e4.png]: /images/20221120/666c8eaf478e44c0ac852c6ce99c01e5.png
[88d9fbf8809f42b027bcf45af44441e0.png]: /images/20221120/bec1d148333f4b49a6bc33a2820aecb7.png
[17731877ef2d44ed40b29870b20a1b79.png]: /images/20221120/9fd64875f5d847de8fd79b57677c633b.png
[58d00d8a4f3ec310cd9cfdf2d9991788.png]: /images/20221120/a419c5269a8d4e9ebf32daee43d6ce40.png
[14b024475aa3cb9b9db553b19862780e.png]: /images/20221120/51112aad0a824a0f90753c297148ccf3.png
[059b8aaf9cd1f9fafdc0b2b9956b8dfe.png]: /images/20221120/dea3ad762ed64ce999f7929dccf39e9e.png
[3169e8c8bcf8ff30e8a8220358d1c89f.png]: /images/20221120/497985ec4ca24f78b3785663be9fbf3f.png
[3131011a97766c607a6224df1c99c37d.png]: /images/20221120/cc0419bd06c44be8a5a2143106591a1e.png