Aerial Lidar Semantic Segmentation Using PointNet++ Deep Learning error when following example

In the "Train Model" phase of the example code,

doTraining = true;
if doTraining
    % Train the network on the ldsTransformed datastore using 
    % the trainNetwork function.
    [net,info] = trainNetwork(ldsTransformed,lgraph,options);
else
    % Load the pretrained network.
    load('pointnetplusTrained','net');
end

The trainNetwork line returns the following error:

Error using trainNetwork

Input datastore returned more than one observation per row for network input 1.

I followed instructions on downloading the DALES dataset and running the code without changing anything, so I'm not sure why I'm getting this error if the example code isn't faulty. Is the problem that I'm running this code on the R2022b release?

Answer 1

[Vinay Ancha]

Hi,

The Aerial Lidar example code and pointnetplusLayers API have been enhanced in R2022b. I belive you might be trying to run the mlx file from R2022a in R2022b and that's the reason the example is running into this error.

Run this command "openExample('deeplearning_shared/AerialLidarSemanticSegmentationUsingPointNetExample')" to open the latest mlx version of this example and try the training from this script.