Function generate_images

gen_videos_interp.py:286–357

Render a latent vector interpolation video. Examples: \b # Render a 4x2 grid of interpolations for seeds 0 through 31. python gen_video.py --output=lerp.mp4 --trunc=1 --seeds=0-31 --grid=4x2 \\ --network=https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/version

(
    network_pkl: str,
    seeds: List[int],
    output: str,
    shuffle_seed: Optional[int],
    truncation_psi: float,
    truncation_cutoff: int,
    grid: Tuple[int,int],
    num_keyframes: Optional[int],
    w_frames: int,
    reload_modules: bool,
    cfg: str,
    image_mode: str,
    sampling_multiplier: float,
    nrr: Optional[int],
    shapes: bool,
    interpolate: bool,
    pose_cond: int,
)

@click.option('--pose_cond', type=int, help='pose_cond angle', default=90, show_default=True)

def generate_images(
    network_pkl: str,
    seeds: List[int],
    output: str,
    shuffle_seed: Optional[int],
    truncation_psi: float,
    truncation_cutoff: int,
    grid: Tuple[int,int],
    num_keyframes: Optional[int],
    w_frames: int,
    reload_modules: bool,
    cfg: str,
    image_mode: str,
    sampling_multiplier: float,
    nrr: Optional[int],
    shapes: bool,
    interpolate: bool,
    pose_cond: int,
):
    """Render a latent vector interpolation video.

    Examples:

    \b
    # Render a 4x2 grid of interpolations for seeds 0 through 31.
    python gen_video.py --output=lerp.mp4 --trunc=1 --seeds=0-31 --grid=4x2 \\
        --network=https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/versions/1/files/stylegan3-r-afhqv2-512x512.pkl

    Animation length and seed keyframes:

    The animation length is either determined based on the --seeds value or explicitly
    specified using the --num-keyframes option.

    When num keyframes is specified with --num-keyframes, the output video length
    will be 'num_keyframes*w_frames' frames.

    If --num-keyframes is not specified, the number of seeds given with
    --seeds must be divisible by grid size W*H (--grid).  In this case the
    output video length will be '# seeds/(w*h)*w_frames' frames.
    """

    print('Loading networks from "%s"...' % network_pkl)
    device = torch.device('cuda:1')
    with dnnlib.util.open_url(network_pkl) as f:
        G = legacy.load_network_pkl(f)['G_ema'].to(device) # type: ignore

    G.rendering_kwargs['depth_resolution'] = int(G.rendering_kwargs['depth_resolution'] * sampling_multiplier)
    G.rendering_kwargs['depth_resolution_importance'] = int(G.rendering_kwargs['depth_resolution_importance'] * sampling_multiplier)
    if nrr is not None: G.neural_rendering_resolution = nrr
    # Specify reload_modules=True if you want code modifications to take effect; otherwise uses pickled code
    # if reload_modules:
    # print("Reloading Modules!")
    G_new = TriPlaneGenerator(*G.init_args, **G.init_kwargs).eval().requires_grad_(False).to(device)
    misc.copy_params_and_buffers(G, G_new, require_all=True)
    G_new.neural_rendering_resolution = G.neural_rendering_resolution
    G_new.rendering_kwargs = G.rendering_kwargs
    G = G_new