Spaces:
Build error
Build error
VikramSingh178
commited on
Commit
•
e2ebd5a
1
Parent(s):
fa32203
chore: Update segmentation model to facebook/sam-vit-huge and adjust prompt and negative prompt for inpainting pipeline
Browse filesFormer-commit-id: b412e4a5cd45780b141847a85bdff7761bc9acc0 [formerly 43c409b17e48b30006ded8ee7797a4db47748837]
Former-commit-id: 7849cb5077ff5dde7887367b0b2eee19f1e96b91
- configs/inpainting.yaml +6 -6
- scripts/pipeline.py +8 -8
configs/inpainting.yaml
CHANGED
@@ -1,13 +1,13 @@
|
|
1 |
-
segmentation_model : 'facebook/sam-vit-
|
2 |
detection_model : 'yolov8l'
|
3 |
model : 'kandinsky-community/kandinsky-2-2-decoder-inpaint'
|
4 |
target_width : 1920
|
5 |
target_height : 1080
|
6 |
-
prompt : '
|
7 |
-
negative_prompt : 'low resolution , bad resolution'
|
8 |
-
roi_scale : 0.
|
9 |
-
strength : 0.
|
10 |
guidance_scale : 7.5
|
11 |
-
num_inference_steps :
|
12 |
output_path : '../outputs'
|
13 |
|
|
|
1 |
+
segmentation_model : 'facebook/sam-vit-huge'
|
2 |
detection_model : 'yolov8l'
|
3 |
model : 'kandinsky-community/kandinsky-2-2-decoder-inpaint'
|
4 |
target_width : 1920
|
5 |
target_height : 1080
|
6 |
+
prompt : 'Award Winning Photography of product on a mountain used for camping'
|
7 |
+
negative_prompt : 'low resolution , bad resolution, bad quality,bad Artifacts,Weird Artifacts , Poor Lighting'
|
8 |
+
roi_scale : 0.6
|
9 |
+
strength : 0.2
|
10 |
guidance_scale : 7.5
|
11 |
+
num_inference_steps : 800
|
12 |
output_path : '../outputs'
|
13 |
|
scripts/pipeline.py
CHANGED
@@ -3,11 +3,11 @@ from diffusers import AutoPipelineForInpainting
|
|
3 |
from diffusers.utils import load_image
|
4 |
from utils import (accelerator, ImageAugmentation, clear_memory)
|
5 |
import hydra
|
6 |
-
from omegaconf import
|
7 |
from PIL import Image
|
8 |
import lightning.pytorch as pl
|
9 |
-
pl.seed_everything(
|
10 |
-
|
11 |
|
12 |
class AutoPaintingPipeline:
|
13 |
"""
|
@@ -26,7 +26,7 @@ class AutoPaintingPipeline:
|
|
26 |
self.image = load_image(image)
|
27 |
self.mask_image = load_image(mask_image)
|
28 |
self.pipeline.to(self.device)
|
29 |
-
|
30 |
|
31 |
|
32 |
def run_inference(self, prompt: str, negative_prompt: str, num_inference_steps: int, strength: float, guidance_scale: float):
|
@@ -36,11 +36,11 @@ class AutoPaintingPipeline:
|
|
36 |
Returns:
|
37 |
Image: The output image after inpainting.
|
38 |
"""
|
39 |
-
|
40 |
image = load_image(self.image)
|
41 |
mask_image = load_image(self.mask_image)
|
42 |
-
output = self.pipeline(prompt=prompt,negative_prompt=negative_prompt,image=image,mask_image=mask_image,num_inference_steps=num_inference_steps,strength=strength,guidance_scale
|
43 |
-
|
44 |
return output
|
45 |
|
46 |
|
@@ -54,7 +54,7 @@ def inference(cfg: DictConfig):
|
|
54 |
"""
|
55 |
augmenter = ImageAugmentation(target_width=cfg.target_width, target_height=cfg.target_height, roi_scale=cfg.roi_scale)
|
56 |
model_name = cfg.model
|
57 |
-
image_path = "../sample_data/
|
58 |
image = Image.open(image_path)
|
59 |
extended_image = augmenter.extend_image(image)
|
60 |
mask_image = augmenter.generate_mask_from_bbox(extended_image, cfg.segmentation_model, cfg.detection_model)
|
|
|
3 |
from diffusers.utils import load_image
|
4 |
from utils import (accelerator, ImageAugmentation, clear_memory)
|
5 |
import hydra
|
6 |
+
from omegaconf import DictConfig
|
7 |
from PIL import Image
|
8 |
import lightning.pytorch as pl
|
9 |
+
pl.seed_everything(1234)
|
10 |
+
|
11 |
|
12 |
class AutoPaintingPipeline:
|
13 |
"""
|
|
|
26 |
self.image = load_image(image)
|
27 |
self.mask_image = load_image(mask_image)
|
28 |
self.pipeline.to(self.device)
|
29 |
+
|
30 |
|
31 |
|
32 |
def run_inference(self, prompt: str, negative_prompt: str, num_inference_steps: int, strength: float, guidance_scale: float):
|
|
|
36 |
Returns:
|
37 |
Image: The output image after inpainting.
|
38 |
"""
|
39 |
+
clear_memory()
|
40 |
image = load_image(self.image)
|
41 |
mask_image = load_image(self.mask_image)
|
42 |
+
output = self.pipeline(prompt=prompt,negative_prompt=negative_prompt,image=image,mask_image=mask_image,num_inference_steps=num_inference_steps,strength=strength,guidance_scale=guidance_scale,height = 1472, width = 2560).images[0]
|
43 |
+
|
44 |
return output
|
45 |
|
46 |
|
|
|
54 |
"""
|
55 |
augmenter = ImageAugmentation(target_width=cfg.target_width, target_height=cfg.target_height, roi_scale=cfg.roi_scale)
|
56 |
model_name = cfg.model
|
57 |
+
image_path = "../sample_data/example5.jpg"
|
58 |
image = Image.open(image_path)
|
59 |
extended_image = augmenter.extend_image(image)
|
60 |
mask_image = augmenter.generate_mask_from_bbox(extended_image, cfg.segmentation_model, cfg.detection_model)
|