Create canny_block.py
Browse files- canny_block.py +87 -0
canny_block.py
ADDED
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import List, Union
|
2 |
+
from PIL import Image
|
3 |
+
import torch
|
4 |
+
import numpy as np
|
5 |
+
|
6 |
+
from diffusers.modular_pipelines import (
|
7 |
+
PipelineState,
|
8 |
+
ModularPipelineBlocks,
|
9 |
+
InputParam,
|
10 |
+
ComponentSpec,
|
11 |
+
OutputParam,
|
12 |
+
)
|
13 |
+
from controlnet_aux import CannyDetector
|
14 |
+
import numpy as np
|
15 |
+
|
16 |
+
|
17 |
+
class CannyBlock(ModularPipelineBlocks):
|
18 |
+
@property
|
19 |
+
def expected_components(self):
|
20 |
+
return [
|
21 |
+
ComponentSpec(name="canny_annotator", type_hint=CannyDetector),
|
22 |
+
]
|
23 |
+
|
24 |
+
@property
|
25 |
+
def inputs(self) -> List[InputParam]:
|
26 |
+
return [
|
27 |
+
InputParam(
|
28 |
+
"image",
|
29 |
+
type_hint=Union[Image.Image, np.ndarray],
|
30 |
+
required=True,
|
31 |
+
description="Image to compute canny filter on",
|
32 |
+
),
|
33 |
+
InputParam(
|
34 |
+
"low_threshold",
|
35 |
+
type_hint=int,
|
36 |
+
default=50,
|
37 |
+
),
|
38 |
+
InputParam("high_threshold", type_hint=int, default=200),
|
39 |
+
InputParam(
|
40 |
+
"detect_resolution",
|
41 |
+
type_hint=int,
|
42 |
+
default=1024,
|
43 |
+
description="Resolution to resize to when running the Canny filtering process.",
|
44 |
+
),
|
45 |
+
InputParam(
|
46 |
+
"image_resolution",
|
47 |
+
type_hint=int,
|
48 |
+
default=1024,
|
49 |
+
description="Resolution to resize the detected Canny edge map to.",
|
50 |
+
),
|
51 |
+
]
|
52 |
+
|
53 |
+
@property
|
54 |
+
def intermediate_outputs(self) -> List[OutputParam]:
|
55 |
+
return [
|
56 |
+
OutputParam(
|
57 |
+
"canny_map",
|
58 |
+
type_hint=Image,
|
59 |
+
description="Canny map for input image",
|
60 |
+
)
|
61 |
+
]
|
62 |
+
|
63 |
+
def compute_canny(self, components, image, low_threshold, high_threshold, detect_resolution, image_resolution):
|
64 |
+
canny_map = components.canny_annotator(
|
65 |
+
input_image=image,
|
66 |
+
low_threshold=low_threshold,
|
67 |
+
high_threshold=high_threshold,
|
68 |
+
detect_resolution=detect_resolution,
|
69 |
+
image_resolution=image_resolution,
|
70 |
+
)
|
71 |
+
return canny_map
|
72 |
+
|
73 |
+
@torch.no_grad()
|
74 |
+
def __call__(self, components, state: PipelineState) -> PipelineState:
|
75 |
+
block_state = self.get_block_state(state)
|
76 |
+
|
77 |
+
block_state.canny_map = self.compute_canny(
|
78 |
+
components,
|
79 |
+
block_state.image,
|
80 |
+
block_state.low_threshold,
|
81 |
+
block_state.high_threshold,
|
82 |
+
block_state.detect_resolution,
|
83 |
+
block_state.image_resolution,
|
84 |
+
)
|
85 |
+
self.set_block_state(state, block_state)
|
86 |
+
|
87 |
+
return components, state
|