Improving the results on real world images

Thanks for this great work!

Any suggestions on how we can improve the performance of the model on real world images? I have seen that almost all of the image matting models perform pretty well on the data which is present in the example images, but don't give good results on actual real world data.