How can a Deep Learning Model be Created for the Extraction of Building Foot Print in Sub Sahara Africa settlement

Question

I am trying to use deep learnig model to extract building foot print in Nigeria settlement, I would be glad to be educated on how to achieve the best model that fit Nigeria settlement setting and pattern .  The available models for America and Europe are not good for the Africa setting.

ShivaniPathak · Answer

Hi @AzeezAfeez,

First you can try pretrained model "Building Footprint Extraction - Africa".

If the pretrained model is not working well for your area of interest and you have building footprint labels for your area you can finetune the pretrained model, you can refer this finetuning guide: https://doc.arcgis.com/en/pretrained-models/latest/imagery/finetuning-the-building-footprint-extraction-africa.htm

Bnewell · Answer

hey @AzeezAfeez have you seen that is also Building Foot Print model for Africa?

* https://www.arcgis.com/home/item.html?id=979cb0cf938946bfb8bb2f41cf9f9795

*Applicable geographies
The model is expected to work in Africa and gives the best results in Uganda and Tanzania.