Why didn’t you pick your own training samples to use in the model? I realized that the model for Africa was based on a particular country which means not for all Africa per se, so I think you need to do some modification to the training data, thank you
Since ESRI built predefined models to use, I wanted to test those. Their help files show strong results for imagery 10-40cm, 8 bit unsigned, three bands, which the open access disaster Maxar data has, so I wanted to see if results were good out of the box, as ESRI said should be. I agree 100%, you have to train to get the best results. I was hoping for good things from the pretrained models, but so far they appear limited. Thanks for you comment.
Incredible content! Thank you for sharing. I will use a similar workflow for tree canopy extraction in the near future. This is super helpful.
Why didn’t you pick your own training samples to use in the model?
I realized that the model for Africa was based on a particular country which means not for all Africa per se, so I think you need to do some modification to the training data, thank you
Since ESRI built predefined models to use, I wanted to test those. Their help files show strong results for imagery 10-40cm, 8 bit unsigned, three bands, which the open access disaster Maxar data has, so I wanted to see if results were good out of the box, as ESRI said should be. I agree 100%, you have to train to get the best results. I was hoping for good things from the pretrained models, but so far they appear limited. Thanks for you comment.
Good work 😄