How should I get a running demo of the pose transformation.

[aliaksandrsiarohin]

How should I get a running demo of the pose transformation.

[AliaksandrSiarohin]

Taichi and fashion is quite limited datasets, so I would not expect them to generalise to arbitrary humans. In your case image look more similar to fashion, try fashion checkpoint and absolute transfer.

[AliaksandrSiarohin]

There is no much difference, just use other checkpoint taichi-cpk.pth.tar or fashion.pth.tar

[eSaniello]

video.zip

Well, I did and here are the results.

This was the code:

import imageio
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.animation as animation
from skimage.transform import resize
from IPython.display import HTML
import warnings
warnings.filterwarnings("ignore")

source_image = imageio.imread('/content/person.jpg')
driving_video = imageio.mimread('/content/taichi.mp4')

source_image = resize(source_image, (256, 256))[..., :3]
driving_video = [resize(frame, (256, 256))[..., :3] for frame in driving_video]

from demo import load_checkpoints
generator, kp_detector = load_checkpoints(config_path='config/taichi-256.yaml', 
                            checkpoint_path='/content/gdrive/My Drive/first-order-model-checkpoints/taichi-cpk.pth.tar')


from demo import make_animation
from skimage import img_as_ubyte

predictions = make_animation(source_image, driving_video, generator, kp_detector, relative=True)

#save resulting video
imageio.mimsave('../generated.mp4', [img_as_ubyte(frame) for frame in predictions])
#video can be downloaded from /content folder

HTML(display(source_image, driving_video, predictions).to_html5_video())

Did I do something wrong?