search

harlanhong / CVPR2022-DaGAN

Official code for CVPR2022 paper: Depth-Aware Generative Adversarial Network for Talking Head Video Generation
https://harlanhong.github.io/publications/dagan.html
Other
958 stars 125 forks source link

Size of input #15

Closed NikitaKononov closed 2 years ago

NikitaKononov commented 2 years ago

Hello Thanks for your great work! I have a question, does your model support input resolution higher, than 256px? 512px for example I see that in code input video and image are resized to 256px, so causes the loss of visual quality Is there a way to use 512x512 img/vid without losing quality?

harlanhong commented 2 years ago

Hi @NikitaKononov,

We only use 256x images for the pressure of computation. If you want to use larger images, you can consider adding several downsampling convolutional layers at the beginning of the network.

NikitaKononov commented 2 years ago

@harlanhong Thanks! So I should retrain model with additional layers? Or adding downsampling layers doesn't require model retraining? I am sorry, if my question is silly :)

harlanhong commented 2 years ago

If you add additional layers, you should retrain the whole network.