A tensorflow implementation of Boundary-Aware Network for Fast and High-Accuracy Portrait Segmentation:https://arxiv.org/pdf/1901.03814.pdf
I'm just trying to re-implement the result in their paper all out of interest, and haven't succeeded. Please be reminded that there might be errors in my code. If anyone have succeeded in re-implementation please let me know.