monodepth作者新论文中的比较:https://arxiv.org/pdf/1806.01260.pdf
KITTI DATASET:
MAKE 3D dataset
1.DDVO
LKVOLearner:Learning Depth from Monocular Videos using Direct Methods, CVPR 2018 不用深度信息做监督,有预训练模型,但是是基于视频帧的
github: https://github.com/MightyChaos/LKVOLearner
2.Laina
Deeper Depth Prediction with Fully Convolutional Residual Networks (FCRN) ,2016 用深度信息做监督,没有训练代码,有网络结构和预训练模型
github: https://github.com/iro-cp/FCRN-DepthPrediction
3.DeepMVS
Learning Multi-View Stereopsis 有训练代码和预训练模型
github: https://github.com/phuang17/DeepMVS
4.DeMoN
Depth and Motion Network 普通照片上效果不错,有代码和预训练模型但是需要固定尺寸,需要验证平行图像上的效果。
github: https://github.com/lmb-freiburg/demon
5.SVSM FT
Single View Stereo Matching 2018 (C++和matlab代码)有预训练模型,caffe编译不过
github: https://github.com/lawy623/SVS
6.ZHAN
Unsupervised Learning of Monocular Depth Estimation and Visual Odometry with Deep Feature Reconstruction 用深度信息做监督,有预训练模型
github: https://github.com/Huangying-Zhan/Depth-VO-Feat
7.Learning Monocular Depth by Distilling Cross-domain Stereo Networks, ECCV18 由于训练模型,看起来像基于monodepth做的研究,前面要装的两个库编译不了,但是要看作者是如何利用真实图片对网络进行微调的。
github:https://github.com/xy-guo/Learning-Monocular-Depth-by-Stereo
8.self-supervised-depth-completion 有预训练模型
github:https://github.com/fangchangma/self-supervised-depth-completion
网友评论