Thin-Plate-Spline-Motion-Model
Thin-Plate-Spline-Motion-Model copied to clipboard
yoyo-nb
Model / Output Resolution?
Is it possible to increase the output resolution or quality? Can the output resolution be increased when building or training the model?
Same question here. I am wondering whether the TPS transform is applicable to high resolution simply by increasing the number of key points and TPS.
Is it possible to increase the output resolution or quality? Can the output resolution be increased when building or training the model?
You can use higher resolution datasets to train models that generate high resolution images (for example, VoxCeleb dataset is 256*256 while TED-talk dataset is 384*384).
However, higher resolution means more memory usage and more time consumption during training. You can also do this with an additional image super-resolution module.
Same question here. I am wondering whether the TPS transform is applicable to high resolution simply by increasing the number of key points and TPS.
The number of keypoints may not be proportional to the result of the motion transfer. In my experiments, 5 keypoints are better than 8 keypoints and 3 keypoints at 256*256 resolution.
More TPS will make motion transfer better, but there's also a limitation.
Same question here. I am wondering whether the TPS transform is applicable to high resolution simply by increasing the number of key points and TPS.
The number of keypoints may not be proportional to the result of the motion transfer. In my experiments, 5 keypoints are better than 8 keypoints and 3 keypoints at 256*256 resolution.
More TPS will make motion transfer better, but there's also a limitation.
256 x 256 pixels is too low of a resolution. 512x512 or 1024x1024 minimum would be ideal. Any chance training has been performed for these resolutions? thanks
seriously in this era 256x256 or 384
could you release 1024x1024 model? at the very least 512x512 please
Are there any training instructions available? I'd love to train a higher resolution model (and share it!)
I want to train 512*512 resolution on the voc dataset. Can someone give me some help, such as croping or filtering the vox dataset to 512 resolution
