The output video have clear box around mouth region.

[rohaantahir]

Hi, I have tried the library its an exceptional work. I am doing inference but facing an issue regarding the output. The output generated have some boxy effect around the mouth. Is there any way to remove it ?

[YoungXu06]

It seems like adjusting resize_factor argument to get a lower resolution video can make the box less obvious，but also blur the whole result video...

[liuquande]

Hi @rohaantahir ,

I also met the same question, have you solved it?

[rohaantahir]

@liuquande no the issue is with their model. It is trained on low resolution that is why bounding box is created.

[NikitaKononov]

You can change the network topology to work with 256x256 or 512x512 resolution and train it from scratch

[wllps1988315]

change img_size from 96 to 256 or 512?

[1800dinky]

change img_size from 96 to 256 or 512?

that just causes errors because the model is expecting 96 so when i change it to 256 it throws the architecture off.

really looking for a fix as the video output quality is pretty bad with that blurry rectangle that appears over the modified portion of the image

[Oyiyi]

Would fine-tune the model resolve this issue? How to do that?

[FreeTymeKiyan]

@1800dinky I'm pretty sure they do it on purpose