Mask-guided

One of the task settings for video subtitle removal in which the algorithm is given the subtitle mask and only needs to inpaint pixels within the specified region. It reflects the upper bound of an inpainting model's capability.