Open kyeno opened 11 months ago
I "worked around" this with a simple bash script additionally invoking ImageMagick:
NOTE: This is made to batch-process entire directories.
#!/bin/sh
for a in *; do
backgroundremover -i "$a" -wn 32 -gb 10 -m "u2net_human_seg" -o "bgr_tmp_$a.png"
convert -flatten "bgr_tmp_$a.png" "bgr_$a.png"
rm -f "bgr_tmp_$a.png"
done
I'll look into this
I have the same question, but I found another solution. I use the model to generate the mask video into a mp4 file. This file is much smaller than the mov file. I use FFmpeg to combine the input video with the mask video on a background image or video to generate the final video into a mp4 file. This way can avoid producing a huge mov file.
Hey!
I tried quite a few variants of using plain-white-image as a background and the
-bi
option, but it doesn't seem to bother at all when working with images (JPEGs in my case).