Open antithing opened 1 year ago
Currently, cannot. This is limited by video inpainting model. STTN has a fair performance but is not scalable to resolution. I'm also looking for a video inpainting model with both good performance and scalability.
Is it just a pixel dimension limitation that STTN has? My input was 1024x512 and 1.9 MB, and then the output was still 1024x512 but only 0.7 MB. I thought the output might look blurrier, but when I went to grab some screenshots I couldn't really tell the difference between the two.
Hi, and thank you for this code!
Is it possible to run remove anything on a video and output the result at the original video resolution (eg, HD / 4kUHD) ? Or is the output downsampled to use the STTN model?
Thanks!