Self-Supervised Learning for Stereo Matching with Self-Improving Ability
2017-12-07 16:48
856 查看
首先呢,这是一个非监督算法,因此它约束的方式就是左右一致性检测,用warp来处理左右图,详见3.1。作者声情并茂的讲述自己就是不要gt。。
网络结构
五部分组成 特征提取 交叉特征向量融合 3D特征匹配 soft-argmin 最后通过图像warp来做约束。
特征提取
有了特征提取,就可以不虚那些复杂区域啦。以往传统的方法相当于是取像素域的原始特征,而深度学习取得特征则是自己去学的
具体特征怎么取的呢,是跟GC-Net一样的。
构建特征向量
通过学习的特征来构建匹配代价空间,需要先对视差范围做一个确定,然后将左右特征图片concated起来,从而生成最终的特征向量
左图复制D份,右图平移D。然后相间的concat。
对3D特征进行regularization
3D卷积反卷积
所以这篇文章就是非监督的GC-Net
简直一模一样~~~
真的是醉了
浪费了时间读这篇文章。。
网络结构
五部分组成 特征提取 交叉特征向量融合 3D特征匹配 soft-argmin 最后通过图像warp来做约束。
特征提取
有了特征提取,就可以不虚那些复杂区域啦。以往传统的方法相当于是取像素域的原始特征,而深度学习取得特征则是自己去学的
具体特征怎么取的呢,是跟GC-Net一样的。
构建特征向量
通过学习的特征来构建匹配代价空间,需要先对视差范围做一个确定,然后将左右特征图片concated起来,从而生成最终的特征向量
左图复制D份,右图平移D。然后相间的concat。
对3D特征进行regularization
3D卷积反卷积
所以这篇文章就是非监督的GC-Net
简直一模一样~~~
真的是醉了
浪费了时间读这篇文章。。
相关文章推荐
- 论文笔记《IS object localization for free? Weakly-supervised learning with convolutional neural network》
- 笔记:Semi-supervised domain adaptation with subspace learning for visual recognition (cvpr15)
- 目标定位--Deep Self-Taught Learning for Weakly Supervised Object Localization
- [Stereo_cnn][cvpr16]Efficient Deep Learning for Stereo Matching(未完成)
- Improved Stereo Matching with Constant Highway Networks and Reflective Confidence Learning
- Efficient Deep Learning for Stereo Matching
- 笔记:Semi-supervised Domain Adaptation with Subspace Learning for Visual Recognition
- Deep Self-Taught Learning for Weakly Supervised Object Localization
- 立体匹配——End-to-End Learning of Geometry and Context for Deep Stereo Regression
- ValueError: "concat" mode can only merge layers with matching output shapes except for the concat
- Stereo Matching文献笔记之(三):《Segment-Tree based Cost Aggregation for Stereo Matching》读后感~
- 论文笔记 A Large Contextual Dataset for Classification,Detection and Counting of Cars with Deep Learning
- 学习《Hardware-Efficient Bilateral Filtering for Stereo Matching》一文笔记。
- Some 3D Graphics (rgl) for Classification with Splines and Logistic Regression (from The Elements of Statistical Learning)(转)
- Policy Gradient Methods for Reinforcement Learning with Function Approximation
- 行人属性“Weakly-supervised Learning of Mid-level Features for Pedestrian Attribute Recognition and Loca”
- Implement regular expression matching with support for '.' and '*'.
- Reading papers_16(Learning hierarchical invariant spatio-temporal features for action recognition with independent subspace analysis)
- Multi-Task Learning with Low Rank Attribute Embedding for Person Re-identification
- 论文笔记之:Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning