How to generate Occupancy Map after registration about deepmapping HOT 2 CLOSED

The general idea is to feed the trained M-Net with the global coordinates to estimate the corresponding occupancy probabilities.

Taking the 2D case as an example, you can
(1) sample a set of points from each point cloud (unoccupied points together with the observed points clouds). You may want to sample a large number of points to cover the environment as much as possible
(2) transform those points into the global frame using the poses estimated from the trained L-Net
(3) use the trained M-Net to estimate the occupancy probabilities
(4) covert the 2D global coordinates to the pixel coordinates and generate the occupancy map from the estimated probabilities and the pixel coordinates.

from deepmapping.

Hi,
Thanks for your quick reply.
I solved this problem now. Actually, this part of function is pretty similar to "self.occp_prob = self.occup_net(inputs)" in your training.

While, on the way to solve this problem. I found sth interesting for your code design, which I will posted it in a different Issues.

this one is closed.

from deepmapping.