图像到BEV转换

Predicting Semantic Map Representations from Images using Pyramid Occupancy Networks提出了一个dense transformer（并非self attention的transformer）的网络结构用于将2D图转换成BEVBird’s-Eye-View Panoptic Segmentation Using Mon

炼丹术师

9346人浏览 · 2021-10-25 19:48:51

炼丹术师 · 2021-10-25 19:48:51 发布

如果用于分割，这个任务又可以叫做Cross-view Semantic Segmentation跨视角语义分割,从第一视角的 2D 图像（First-view Observation）得到俯视语义图（Top-down-view Semantic Map）的过程称作跨视角语义分割．如果用于检测，那就是bev感知。

Predicting Semantic Map Representations from Images using Pyramid Occupancy Networks(PON)

提出了一个dense transformer（并非self attention的transformer，只是MLP结构）的网络结构用于将2D图转换成BEV
https://githubhelp.com/tom-roddick/mono-semantic-maps
code:https://github.com/tom-roddick/mono-semantic-maps
在这里插入图片描述

 https://githubhelp.com/tom-roddick/mono-semantic-maps

Bird’s-Eye-View Panoptic Segmentation Using Monocular Frontal View Images

用到了上一篇方法中的dense transformer
https://zhuanlan.zhihu.com/p/415109656
项目主页: http://panoptic-bev.cs.uni-freiburg.de/
code:https://github.com/robot-learning-freiburg/PanopticBEV
在这里插入图片描述

Cross-View Semantic Segmentation for Sensing Surroundings(VPN)

在这里插入图片描述

Orthographic Feature Transform for Monocular 3D(OFT)

https://zhuanlan.zhihu.com/p/60117934
在这里插入图片描述

Lift, Splat, Shoot: Encoding Images From Arbitrary Camera Rigs by Implicitly Unprojecting to 3D

code : https://github.com/nv-tlabs/lift-splat-shoot.git 在这里插入图片描述

MonoLayout: Amodal Scene Layout from a single image

在这里插入图片描述

code:https://github.com/hbutsuak95/monolayout?fileGuid=3X8QJDGGJPXyQgW9

Projecting Your View Attentively: Monocular Road Scene Layout Estimation via Cross-view Transformation

code:https://github.com/JonDoe-297/cross-view
在这里插入图片描述

Structured Bird’s-Eye-View Traffic Scene Understanding from Onboard Images

在这里插入图片描述
https://github.com/ybarancan/STSU

Translating Images into Maps

在这里插入图片描述
https://github.com/avishkarsaha/translating-images-into-maps
和Predicting Semantic Map Representations from Images using Pyramid Occupancy Networks思想差不多，图像竖直方向有全局深度信息．