Annotated Examples

i = Image.open(image_path)
i = ImageOps.exif_transpose(i)
if i.mode == 'I':
    i = i.point(lambda i: i * (1 / 255))
image = i.convert("RGB")
image = np.array(image).astype(np.float32) / 255.0
image = torch.from_numpy(image)[None,]

for (batch_number, image) in enumerate(images):
    i = 255. * image.cpu().numpy()
    img = Image.fromarray(np.clip(i, 0, 255).astype(np.uint8))
    filepath = # some path that takes the batch number into account
    img.save(filepath)

mask = 1.0 - mask

# We want [B,H,W,C] with C = 1
if len(mask.shape)==2: # we have [H,W], so insert B and C as dimension 1
    mask = mask[None,:,:,None]
elif len(mask.shape)==3 and mask.shape[2]==1: # we have [H,W,C]
    mask = mask[None,:,:,:]
elif len(mask.shape)==3:                      # we have [B,H,W]
    mask = mask[:,:,:,None]

# Invert mask back to original transparency layer
mask = 1.0 - mask

# Unsqueeze the `C` (channels) dimension
mask = mask.unsqueeze(-1)

# Concatenate ("cat") along the `C` dimension
rgba_image = torch.cat((rgb_image, mask), dim=-1)

class Noise_MixedNoise:
    def __init__(self, nosie1, noise2, weight2):
        self.noise1  = noise1
        self.noise2  = noise2
        self.weight2 = weight2

    @property
    def seed(self): return self.noise1.seed

    def generate_noise(self, input_latent:torch.Tensor) -> torch.Tensor:
        noise1 = self.noise1.generate_noise(input_latent)
        noise2 = self.noise2.generate_noise(input_latent)
        return noise1 * (1.0-self.weight2) + noise2 * (self.weight2)