Detect Object Using Deep Learning. Error 'No module named torchvision.models.detection.retinanet'

[MattiaScalas]

I'm trying to use the TreeDetection deep learning model with some raster imagery (ArcGIS Pro 2.8).

I get an error 002667 Cannot load python module when I add TreeDetection.dlpk as Model Definition in the geoprocessing Detect Object Using Deep Learning. The error is:

ModuleNotFoundError: No module named 'torchvision.models.detection.retinanet'

How can I solve this problem? Is it a problem linked to my ArcGIS version?

Thanks

[DanPatterson]

were you following

Train Deep Learning Model (Image Analyst)—ArcGIS Pro | Documentation

for your version of Pro?

[PavanYadav]

@MattiaScalas Have you installed the Deep Learning Packages? if not, please download and install them (see link below) and retry. 

https://github.com/Esri/deep-learning-frameworks

[ShariF]

Hi. 

I was having the same issue. I was working in a cloned environment under Package Manger and kept getting errors. I realized when I read the fine print of installing the deep learning frameworks, that the installation from github is updating the non-cloned environment. I had always read to NOT touch the arcgispro-py3 environment or there would be dire consequences. I do not get errors now. Good luck and I hope you have success with the Tree Detection. I'm trying out the Tree Segmentation for the first time.