U-2-Net
U-2-Net copied to clipboard
Is it possible to create conditional version of model?
The model performs great on a single frame. But I am trying to apply this to video and the result is kind of flickering. I would like to know is it possible to update the model to provide it with 4 channel input where 4-th channel is the mask of the previous frame? Would it be beneficial? Maybe there is such an implementation somewhere already?