OSM Import

Accelerated Computing Aerial Forum
1

Hi,

I have just got up and running with AODT and would like to import a custom OSM scene.

I have been following this guide: Scene Importer - NVIDIA Docs but have been experiencing issues.

I think the issue is in the container setup, as is described first in the guide. However, without this step, I was able to connect a worker in the GIS tab and run the GIS processing (in theory), which the guide suggests would not be possible. In the UI it shows success but the backend displays an error, as shown in the image. This and the USD that is generated being empty makes me think it is due to being unable to do the container setup.

image
image1365×527 141 KB

When I try to run channel_listener.py from within nvcr.io/esee5uzbruax/aodt-gis:1.1.0, I find that the IP of my nucleus server (on the same machine/ localhost) is not reachable. Results of trying this and the CLI script attached. I also cannot test the networking with ping from the container.

image
image669×469 132 KB
I

Thanks in advance for your help in fixing this.

2

@oed
When you bring up the container, can you please add the following flag and retry?
–network=host

e.g. docker run --rm -it --network=host -v <local_path>:/src/aodt_gis/data --pull=always nvcr.io/esee5uzbruax/aodt-gis:1.1.0 /bin/bash

3

Additionally, the latitudes are flippled. Please run this:
python3 gis_jobs/osm_job.py -o /src/aodt_gis/data/bristol_1.usd -c -2.612 51.443 -2.57 51.46

4

Hi,

Thanks for your response. I have now successfully imported my scene using the UI. But only by specifying in lower-left longitude, lower left latitude, upper-right longitude, upper-right latitude, as shown in the image. This is the opposite to the ? prompt shown in the AODT window. For repeatability, this only worked once I had restarted the backend from CLI and ran the nvcr.io/esee5uzbruax/aodt-gis:1.1.0 container to successfully attach a gis worker.

image
image1078×294 23 KB

When trying via the CLI approach however, I didn’t have any success - I could run the container, with --network=host but still could not run the .py file as it could not see my nucleus server.

I have solved my issue via the UI but I just wanted to leave a detailed description of what was unclear.

Thanks for your help.