video

[GuoQuanhao]

python main.py -i video_folder -m mean_face.npy -o out_folder -l landmarks_file -w smoothing_window -s scale --offset 0.11 0.336 0.155 usage: main.py [-h] [--scale SCALE] [--smoothing_window SMOOTHING_WINDOW] [--file_list [FILE_LIST]] [--offset OFFSET [OFFSET ...]] [--input INPUT] [--landmarks LANDMARKS] [--output OUTPUT] [--subjects SUBJECTS [SUBJECTS ...]] [--mean [MEAN]] [--calculate_mean] [--gpu] [--visualise [VISUALISE]] [--ext_video EXT_VIDEO] [--append_extension APPEND_EXTENSION] [--add_extension] [--picture] [--append_folder] [--reference REFERENCE] [--workers WORKERS] [--out_size OUT_SIZE [OUT_SIZE ...]]

I don't know the command, what is smoothing_window, what is scale,what is offset?

[DinoMan]

The smoothing window will use the average landmarks over a window. Since landmark detection is a bit jittery setting a higher smoothing window will result in videos with less jitter. When you provide a mean face it has no borders (i.e. the frame fits only the face). If you want to add some margins you can do this with the offset option (side top bottom). The offset is provided in percentage of the mean face dimensions. The scale option scales your mean face.

I haven't found the time to properly document this repo but I hope to do it in the near future.