Object detection on oriented imagery

[dkutkevicius_hb]

Hello,

Have anyone tried to use Detect Object Using Deep Learning GP tools on Oriented Imagery layer? This is a new capability according to this blog post. I have tried to test it with TextSAM dlpk, objects are detected, but does not have any geometry.

At first I thought that maybe TextSAM does not detect the objects I am asking for (cars), so I tried to run it directly on the source image - it worked, cars were detected. But when I use the same image in oriented imagery catalog, detected cars are not recorded. See attached images for visuals. 

[ToddW_stl]

I haven't (yet), but i'm interested in trying it now.  Have you reached out to the authors of that blog post or the developers (since you're Esri)?  My initial thought is a projection issue/mismatch, or something to do with duplicate features detecting since the tool doc says:

"Use the Non Maximum Suppression parameter to identify and remove duplicate features from the object detection. To learn more about this parameter, see the Usage section of the Non Maximum Suppression tool. When the inputs are oriented imagery layers, the duplicates are retained with null ground geometries."

@GregoryBrunner - have you worked with GeoAI and OI yet?

[dkutkevicius_hb]

I have seen this parameter and it got my attention since it involves objects with null geometries, which are exactly what I am getting. I have tried to enable / disable this option during processing, but it didn't change the output, results are the same either way.

I have noticed that objects are detected if 360 degree oriented imagery dataset is used as a input (example dataset: https://www.arcgis.com/home/item.html?id=c8ba6ef571b840979ad6a254bf3cf7d8). But if its oblique imagery, null geometries are returned.