Detect Objects using Deep Learning Error When Using GPU Error 9999999 and Parallel Processing Job Timed Out

[MayarAlzerki]

Hello everyone,

I am using a pretrained model for building detection on ArcGIS 3.0.0 Pro, when I set the processor type as CPU it works perfectly, but I want to use the GPU which is throwing these errors:

My GPU is Nvidia GeForce RTX 2070 Super. I checked if torch.cuda.is_available() returns true or not and it returned True.

Please help me in this.

Thank you in advance,

[PavanYadav]

@MayarAlzerki 

Can you share the name of the pre-trained model? Are you using a local raster (e.g tif , jpg) or using image service or webgis layer as an input raster? 

regards

Pavan 

[MayarAlzerki]

Hello Pavan,

Thank you for your response. The name of the pre-trained model is: Building Footprint Extraction - KSA.

Please find it through the following link: https://esrisaudi-ai.maps.arcgis.com/home/item.html?id=63785ee9860d4bd68cd82caccdcf1742

I'm using the world imagery service layer.

Thank you in advance,
Mayar

[AderonkeAdetoro]

Hello @PavanYadav I encountered the same problem with a tiff imagery I acquired using drone.

I couldn't use my dedicated GPU, Nvidia with 4gb but it ran with the other GPU but the result wasn't satisfactory, I got a result with each building having multiple footprints. Then when I trained my model and tried to run it again, I could only use CPU which made my processing time really slow and also with unsatisfactory result. Could you please help me with a solution?  The USA Footprint detected some of my buildings in Nigeria but with multiple footprints for a building and smaller footprint size compared to the actual building size for other buildings.

[PavanYadav]

@MayarAlzerkiI am sorry, I missed your last comment. When you use the World Imagery basemap layer, you need to define an extent (try a small one in the environment tab) and a cell size (something between 0.1 and 0.5). I agree, the error message was wrong. However, my understanding is that there are some licensing limitations on how you can use the World Imagery basemap as an input.

@VinayViswambharan can you comment on the usages of the world imagery for AI inferencing? 

thanks
Pavan

[MayarAlzerki]

Hello @PavanYadav,

It's fine no worries, actually, I'm not working on it anymore, but on my free time I would give it a try for sure!

Appreciate your response. Thank you!

[VinayViswambharan1 · Accepted solution]

Hi @MayarAlzerki - The world imagery cannot be used for inferencing or programmatic access. It is violating the terms of use of the basemap (which is designed currently to be used only as a basemap).

Regards,

Vinay

[MayarAlzerki]

Hello @VinayViswambharan1!

Apologiess, I think that I did not get your point. Do you mean that it's not allowed in GPU only? Because I can do detection using my CPU

[PavanYadav]

@MayarAlzerki it's a license limitation.