Abhishek Singh, a young Indian inventor, is conducting research on erasing a smartphone reflected in a mirror through a neural network. Usually, when taking a selfie, they also capture what they see through the bedroom or bathroom mirror. In this case, you are trying to erase the smartphone that is visible on the screen.
This structure suggests that the learned neural network detects the smartphone reflected in the mirror, draws a border around it, and then hangs a mask. It is said that 18,000 images were used for learning. It is said that this technology uses deep learning and Google Tensorflow, a multi-layer neural network library, and uses two neural networks. The library I used is Keras.
The developer said that while viewing selfie photos around the world, it was interesting that it would be fun to just remove the smartphone, and this technology was developed. Related information can be found here .