I think i find the way to change the landmark metadata in each jpg file to match the src face.but base on my skill i probably have to code fulltime for a week to achieve that.And im afraid it will be ignore in the merger process.
In the merger.py i see this LandmarksProcessor.get_transform_mat and LandmarksProcessor.transform_points.I have no idea how each work but it sound like it will recalculate landmark postions again.
I think i find the way to change the landmark metadata in each jpg file to match the src face.but base on my skill i probably have to code fulltime for a week to achieve that.And im afraid it will be ignore in the merger process. In the merger.py i see this LandmarksProcessor.get_transform_mat and LandmarksProcessor.transform_points.I have no idea how each work but it sound like it will recalculate landmark postions again.