feat(api): blend outpainting border with noise

api/onnx_web/serve.py

@@ -103,22 +103,30 @@ def get_latents_from_seed(seed: int, width: int, height: int) -> np.ndarray:
 
 
 # based on https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/master/scripts/outpainting_mk_2.py#L175-L232
-def expand_image(source_image: Image, dims: Tuple[int, int, int, int]):
+def expand_image(source_image: Image, mask_image: Image, dims: Tuple[int, int, int, int]):
     (left, right, top, bottom) = dims
 
     full_width = left + source_image.width + right
     full_height = top + source_image.height + bottom
 
-    full_image = Image.new('RGB', (full_width, full_height), 'white')
-    full_image.paste(source_image, (left, top))
+    full_source = Image.new('RGB', (full_width, full_height), 'white')
+    full_source.paste(source_image, (left, top))
 
-    # draw = ImageDraw.Draw(full_image)
-    # draw.rectangle((0, 0, left, full_height), 'white')
-    # draw.rectangle((0, 0, full_width, top), 'white')
-    # draw.rectangle((full_width - right, 0, full_width, full_height), 'white')
-    # draw.rectangle((0, full_height - bottom, full_width, full_height), 'white')
+    full_mask = Image.new('RGB', (full_width, full_height), 'white')
+    full_mask.paste(mask_image, (left, top))
 
-    return full_image
+    full_noise = Image.effect_noise((full_width, full_height), 2)
+
+    for x in range(full_source.width):
+        for y in range(full_source.height):
+            mask_color = full_mask.getpixel((x, y))
+            noise_color = full_noise.getpixel((x, y))
+            source_color = full_source.getpixel((x, y))
+
+            if mask_color[0] > 0:
+                full_source.putpixel((x, y), source_color + (mask_color * noise_color))
+
+    return (full_source, full_mask, (full_width, full_height))
 
 
 def load_pipeline(pipeline: DiffusionPipeline, model: str, provider: str, scheduler):
@@ -288,9 +296,8 @@ def run_inpaint_pipeline(model, provider, scheduler, prompt, negative_prompt, cf
     latents = get_latents_from_seed(seed, width, height)
     rng = np.random.RandomState(seed)
 
-    full_mask = expand_image(mask_image, (256, 256, 256, 256))
-    full_source = expand_image(source_image, (256, 256, 256, 256))
-    # full_source.thumbnail((512, 512))
+    extra = 256
+    (full_source, full_mask, full_dims) = expand_image(source_image, mask_image, (extra, extra, extra, extra))
 
     image = pipe(
         prompt,