Hello, I used the "Detect Objects Using Deep Learning" tool to identify individual building. However, the result is not entirely satisfactory. Some buildings are nicely identified, others have more that one envelop/footprint, and a few are missed. I changed the parameters one way or another which improved the outcome but it's still not perfect. Any idea how to fix this?
Thanks.
Hi @NHM Are you using one of the pre-trained models? Please note these models have specific image resolutions requirements for example the Building Footprint Extraction - USA model works well with high-resolution (10–40 cm) imagery. In the Detect Objects using Deep Learning tool you can try decreasing the threshold and/or enhancing the resolution of your input raster.
Hope this helps!