Modern way to estimate homography matrix(by light weight cnn)

I improve the results introduce by the paper Deep image homography estimation. I write the details and post the link of the trained model at here.

Following are some of the results, the region estimated by the model(red rectangle) is very close to the real regions(blue rectangle).

Unfortunately, this model do not work well for real world data, I guess this is why the paper never mentioned how to do it.