qsisi
qsisi
> I think such a mask can be implemented by modifying the [NIDCost](https://github.com/koide3/direct_visual_lidar_calibration/blob/e110ca230e9e262a93d7b2f90c2d498dbc970e71/include/vlcal/costs/nid_cost.hpp#L23) class that measures the normalized information distance between a point cloud and an image. > > BTW,...
> Sky region depth estimation is really hard for our model. This is because our training data do not enclose much such supervision. I recommend using a semantic segmentation model...
Looks like something goes wrong with the variable **coors**. Maybe you could consider printing it out for further debugging.
Looking like changing an open3d version could solve the problem.
> I used open3d0.16/0.18 to get this error https://github.com/zhulf0804/GCNet/blob/main/requirements.txt#L6
> Hi for the depth prediction, it is okay to use relu although we found exp generally works better. For the point cloud prediction, since point can be (-inf, inf),...