Hello, I am new to Arc GIS Pro and I want to use it for detecting objects in my GeoTIFF Map using my own trained model using PyTorch 1.12 (.pth extention). Is it possible for me to convert my .pth model to dlpk so I can use it on ArcGIS Pro Deep Learning Package?
You have two options either you train the model with arcgis pro itself (https://developers.arcgis.com/python/samples/automate-building-footprint-extraction-using-instance-s... ) or if you still want to your own model then you have to ditch the arcgis pro environment and just use python to get the result.
There should be a way as we get .pth and .dlpk file when training your own model in Arcgis Environment. ArcGIS object detection/image classification is build on top of pytorch framework thats why we get .pth file, means there's a post-processing added after .pth file is genrated which would require more digging.
@Avid Please see - https://github.com/Esri/raster-deep-learning?rsource=https%3A%2F%2Flinks.esri.com%2FRasterDL
Please try - https://github.com/Esri/raster-deep-learning?rsource=https%3A%2F%2Flinks.esri.com%2FRasterDL