False positives - where to report?

Hi,
Very cool project. In many cases this works and are a cool resource. But unfortunately around where I live in Sweden there are to many examples of buildings detected that are in the ocean, on cliffs/bare rocks, just shadows in the forrest etc.

Is there a plan to rebuild this data set so you are interested in getting feedback on what has worked or does not work?

I added four examples below that are all within 5 km distansce from each other. Non of the marked buildings as seen in the images are actual buildings. The satelite imagery shown as background are bing satelite imagery. http://ecn.t3.tiles.virtualearth.net/tiles/a{q}.jpeg?g=1, which is stated to be bing virtual earth. I am not entirely sure if that is the imagery used but anyhow - in real life there are no buildings in the places where this project has indicated there is one.

Start by reviewing the four linked examples and the repository's guidance for reporting building-footprint errors. Determine whether the false positives are dataset issues or model-output issues, then define a feedback or correction process; no source file or test is identified in the issue.