01-24-2019 06:55 AM
The ‘Dedicated GPU memory usage’ graph monitors the GPU memory usage by applications. The graph shows that 5GB of 6GB available is being used. This also explains why 2GB video memory is not sufficient for the operator to run under.
03-28-2019 07:14 AM
I'm trying to use Initialize Object Detection and I have some problem.
I run the procedure using 10 Training Steps and always end with successful. After that, I tried to run again the same model but with 100 Traing Steps. Unfortunately, the Spatial model execution failed.
Here, you can see the Session Log:
28/03/19 14:18:49 SessionMgr(1356): Executing spatial model
28/03/19 14:18:49 SessionMgr(1356): Running command line: C:/Program Files/Common Files/Hexagon/SpatialModeler/16.5/bin/x64URelease/mlpywrapper_gpu.exe "C:/Program Files/Hexagon/ERDAS IMAGINE 2018/etc/MachineLearning/Operators/initializeobjectdetection.
py" "C:/Users/utente/AppData/Local/Temp/SPATIAL_MODELER-82fb-484e-09c2-bac3-011516/mlpy_jsoninput_492b9140-dbcb-4be9-ba0c-bb1093451a7a.Json" "C:/Users/utente/AppData/Local/Temp/SPATIAL_MODELER-82fb-484e-09c2-bac3-011516/mlpy_jsonoutput_254f6596-e877-47f
28/03/19 14:25:50 SessionMgr(1356): [0, 900000, 1200000]
28/03/19 14:25:50 SessionMgr(1356): Converted 530 variables to const ops.
28/03/19 14:25:50 SessionMgr(1356): Image with id b'ab70cb0fdc2884fd6d51ff531e577ba919d99b8f0e2d46883cac7884acc6bba3' already added.
28/03/19 14:25:53 SessionMgr(1356): Could not understand JSON for type: IMAGINE.File
28/03/19 14:25:53 SessionMgr(1356): Spatial model execution failed.
Where is the problem?
Thanks for your help,
03-28-2019 09:08 AM - edited 03-28-2019 09:56 AM
Are you using your GPU to run the initialization? If so, I suspect you may be running out of GPU Memory.
You may turn off the GPU in the preference and run the initialization. It will take longer but will run to completion.
03-29-2019 10:14 AM
It's not solved :-)
We are having problems also whan working on the CPU. The behaviour of these operators is still quite unstable and we have been spending a lot of time to make them working on. My colleague Nicolò will add other info about it.
04-05-2019 09:18 AM
We have just finished a training for our customers.
We created the miz file on our machine (before the training) and then we built with the students the model for the object detection.
Every person had the same demo raster to classifiy (a little subset with some airplanes).
The model stopped on 4 computers on a total number of 32 computers. Everyone was working on CPU (cuda not enabled, restart erdas imagine).
We had the same behaviour while testing the model, before the training. On some computer it stops and we do not know why.