Implement LatentSync

[ooe1123]

#1637

[kyakuno]

unetが4.41GB。
https://drive.google.com/drive/u/1/folders/1TDsNd9UwXhK41R1d7diSrIGEmxqIz5w6

[kyakuno]

モデルアップロード済み。
https://storage.googleapis.com/ailia-models/latentsync/unet.onnx

[kyakuno]

macOSで下記のエラーになる。

  File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/diffusion/latentsync/latentsync.py", line 330, in <module>
    main()
  File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/diffusion/latentsync/latentsync.py", line 326, in main
    recognize_from_video(pipe)
  File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/diffusion/latentsync/latentsync.py", line 197, in recognize_from_video
    ) = affine_transform_video(video_path, size=256)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/diffusion/latentsync/latentsync.py", line 166, in affine_transform_video
    image_processor = ImageProcessor(size=size)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/diffusion/latentsync/image_processor.py", line 52, in __init__
    self.fa = face_alignment.FaceAlignment(
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.12/site-packages/face_alignment/api.py", line 80, in __init__
    self.face_detector = face_detector_module.FaceDetector(device=device, verbose=verbose, **face_detector_kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.12/site-packages/face_alignment/detection/sfd/sfd_detector.py", line 31, in __init__
    self.face_detector.to(device)
  File "/opt/homebrew/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1173, in to
    return self._apply(convert)
           ^^^^^^^^^^^^^^^^^^^^
  File "/opt/homebrew/lib/python3.12/site-packages/torch/nn/modules/module.py", line 779, in _apply
    module._apply(fn)
  File "/opt/homebrew/lib/python3.12/site-packages/torch/nn/modules/module.py", line 804, in _apply
    param_applied = fn(param)
                    ^^^^^^^^^
  File "/opt/homebrew/lib/python3.12/site-packages/torch/nn/modules/module.py", line 1159, in convert
    return t.to(
           ^^^^^
  File "/opt/homebrew/lib/python3.12/site-packages/torch/cuda/__init__.py", line 284, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

[kyakuno]

face_alignmentをblazefaceに置き換えるオプションを追加したい。

[kyakuno]

FaceAlignmentに下記を指定はできる。

            device='cpu'