Train building classifier for imagerial data


I am using the api for accessing imagery of aerial data.

I would like to train a model that can caputure on the imagery if the provided street address has a house, pool, garage etc.?

Any suggestions how to start such a project? Any recommendations for similar implementations?

Looking forward to your replies!


Posted 2020-02-20T19:09:09.930

Reputation: 165



This blog article might be a good starting point. From what you described and depending on your data, semantic segmentation might be overkill and classification will suffice.

Either way, the first step will be to get your hands on training data. If you do not have labels already, this might mean that you have to sit down for a while an label a bunch of images. This is much easier for classification than segmentation, but the information you end up with is not as rich.

For segmentation you should then check out U-Net or DeepLabv3. Classification is not my strong site, but VGG or a ResNet should do the trick.


Posted 2020-02-20T19:09:09.930

Reputation: 1 409