search

SimonTheVillain / ScalableFusion

Completely mesh based RGB-D reconstruction. Capable of storing large reconstructions without wasting GPU memory.
GNU General Public License v3.0
77 stars 12 forks source link

/usr/bin/ld: cannot find -lopencv_cvv #4

Open Quitino opened 4 years ago

Quitino commented 4 years ago

thank you for you work! when I build it, an error occurred:/usr/bin/ld: cannot find -lopencv_cvv.My system configuration：

Hope to get your help, thank you very much.

Quitino commented 4 years ago

I guess it's an opencv version problem, so I use opencv4.Maybe I did not change the opencv version successfully. Because the result of cmake is shown in the figure, no matter how I change the CMakeLists.txt file, it will always find opencv that comes with ROS, and this version is not compiled cuda10.1. Hope to get your help, thank you very much.

2019-12-07 16-59-46屏幕截图

SimonTheVillain commented 4 years ago

Hi Quitino, I honestly can't tell you why cmake is referring to the kinetic version of OpenCV. I can tell you though, that we dropped the requirement of OpenCV being compiled with CUDA. I also found it way easier compiling that thing with Ubuntu 18.04

This library is under heavy restructuring and I would not advise to use it at all. Currently it is not very stable, neither on the main branch, nor on any of the development branches. Since we are only utilising depth for odometry, there is only very brittle camera tracking. Approaches like ElasticFusion utilise photometric approaches as well, so for a better experience you might want to look into these libraries.