lsd-slam test
some words before...
semantics are necessary to build bigger and better SLAM systems.
Robotics is the killer application of SLAM
Will end-to-end learning soon replace the mostly manual labor involved in building today’s SLAM systems?.
-> Integrating semantics into SLAM is often talk about, but it is easier said than done. Moreno's PhD thesis: Dense Semantic SLAM
SLAM (construction) <---help---> Deep Learning (perception) large-scale "correspondence engines" -> large-scale datasets
调试过程:
1. 确认摄像头是或否工作,运用cheese
2. 安装ros下的库uvc_camera,确认摄像头所对应设备号/dev/video*
3. 在uvc_camera下新建launch文件夹,再新建uvc_camera_node.launch文件,复制以下代码