My project is to use deep learning, essentially a UNET segmentation model, to detect impervious surfaces on high resolution aerial photographies.

I wonder if it's better to train the model with many different impervious surfaces classes like buildings, roads, parking lots, cars, etc... or if using a single class that regroup each of them could work the same?

I have to create my labels and masks myself so using a single class without differentiating the categories of impervious surface would be easier and will take less time. However, precision is what matter the most.

Any suggestions? up



You must log in to answer this question.