Spaces:
Runtime error
Runtime error
VikramSingh178
commited on
Commit
•
6e67c16
1
Parent(s):
6b857b8
commit
Browse filesFormer-commit-id: 522ceaeca59f5a0db3c0e1e300b6ac98b69656af
- __pycache__/config_settings.cpython-310.pyc +0 -0
- api/models/__pycache__/__init__.cpython-310.pyc +0 -0
- api/models/__pycache__/sdxl_input.cpython-310.pyc +0 -0
- api/routers/__pycache__/painting.cpython-310.pyc +0 -0
- api/routers/__pycache__/sdxl_text_to_image.cpython-310.pyc +0 -0
- api/routers/painting.py +50 -34
- api/routers/sdxl_text_to_image.py +1 -1
- api/yolov8l.pt.REMOVED.git-id +1 -0
- picpilot.egg-info/SOURCES.txt +3 -2
- picpilot.egg-info/top_level.txt +1 -0
- sample_data/mask_image.png +0 -0
- scripts/__pycache__/__init__.cpython-310.pyc +0 -0
- scripts/__pycache__/api_utils.cpython-310.pyc +0 -0
- scripts/__pycache__/config.cpython-310.pyc +0 -0
- scripts/__pycache__/inpainting_pipeline.cpython-310.pyc +0 -0
- scripts/__pycache__/s3_manager.cpython-310.pyc +0 -0
- scripts/api_utils.py +2 -2
- scripts/inpainting_pipeline.py +1 -1
- scripts/kandinsky3_inpainting.py +10 -10
- scripts/load_pipeline.py +1 -1
- ui/__pycache__/ui.cpython-310.pyc +0 -0
- ui/ui.py +38 -109
__pycache__/config_settings.cpython-310.pyc
ADDED
Binary file (598 Bytes). View file
|
|
api/models/__pycache__/__init__.cpython-310.pyc
CHANGED
Binary files a/api/models/__pycache__/__init__.cpython-310.pyc and b/api/models/__pycache__/__init__.cpython-310.pyc differ
|
|
api/models/__pycache__/sdxl_input.cpython-310.pyc
CHANGED
Binary files a/api/models/__pycache__/sdxl_input.cpython-310.pyc and b/api/models/__pycache__/sdxl_input.cpython-310.pyc differ
|
|
api/routers/__pycache__/painting.cpython-310.pyc
CHANGED
Binary files a/api/routers/__pycache__/painting.cpython-310.pyc and b/api/routers/__pycache__/painting.cpython-310.pyc differ
|
|
api/routers/__pycache__/sdxl_text_to_image.cpython-310.pyc
CHANGED
Binary files a/api/routers/__pycache__/sdxl_text_to_image.cpython-310.pyc and b/api/routers/__pycache__/sdxl_text_to_image.cpython-310.pyc differ
|
|
api/routers/painting.py
CHANGED
@@ -1,22 +1,20 @@
|
|
1 |
-
from pathlib import Path
|
2 |
import os
|
3 |
import uuid
|
4 |
from typing import List, Tuple, Any, Dict
|
5 |
-
from fastapi import APIRouter, File, UploadFile, HTTPException, Form
|
6 |
from pydantic import BaseModel, Field
|
7 |
from PIL import Image
|
8 |
import lightning.pytorch as pl
|
9 |
from scripts.api_utils import pil_to_s3_json, pil_to_b64_json, ImageAugmentation, accelerator
|
10 |
-
from inpainting_pipeline import AutoPaintingPipeline, load_pipeline
|
11 |
from hydra import compose, initialize
|
12 |
from async_batcher.batcher import AsyncBatcher
|
13 |
import json
|
14 |
from functools import lru_cache
|
15 |
-
import asyncio
|
16 |
pl.seed_everything(42)
|
17 |
router = APIRouter()
|
18 |
|
19 |
-
|
20 |
with initialize(version_base=None, config_path="../../configs"):
|
21 |
cfg = compose(config_name="inpainting")
|
22 |
|
@@ -31,7 +29,6 @@ def load_pipeline_wrapper():
|
|
31 |
"""
|
32 |
pipeline = load_pipeline(cfg.model, accelerator(), enable_compile=True)
|
33 |
return pipeline
|
34 |
-
|
35 |
inpainting_pipeline = load_pipeline_wrapper()
|
36 |
|
37 |
class InpaintingRequest(BaseModel):
|
@@ -45,7 +42,8 @@ class InpaintingRequest(BaseModel):
|
|
45 |
guidance_scale: float = Field(..., description="Guidance scale for inference")
|
46 |
mode: str = Field(..., description="Mode for output ('b64_json' or 's3_json')")
|
47 |
num_images: int = Field(..., description="Number of images to generate")
|
48 |
-
|
|
|
49 |
class InpaintingBatchRequestModel(BaseModel):
|
50 |
"""
|
51 |
Model representing a batch request for inpainting inference.
|
@@ -68,14 +66,35 @@ async def save_image(image: UploadFile) -> str:
|
|
68 |
f.write(await image.read())
|
69 |
return file_path
|
70 |
|
71 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
72 |
"""
|
73 |
Run inference using an inpainting pipeline on an image.
|
74 |
|
75 |
Args:
|
76 |
cfg (dict): Configuration dictionary.
|
77 |
image_path (str): Path to the image file.
|
78 |
-
mask_image_path (str): Path to the mask image file.
|
79 |
request (InpaintingRequest): Pydantic model containing inference parameters.
|
80 |
|
81 |
Returns:
|
@@ -84,8 +103,17 @@ def run_inference(cfg, image_path: str, mask_image_path: str, request: Inpaintin
|
|
84 |
Raises:
|
85 |
ValueError: If an invalid mode is provided.
|
86 |
"""
|
87 |
-
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
painting_pipeline = AutoPaintingPipeline(
|
90 |
pipeline=inpainting_pipeline,
|
91 |
image=image,
|
@@ -107,33 +135,26 @@ def run_inference(cfg, image_path: str, mask_image_path: str, request: Inpaintin
|
|
107 |
raise ValueError("Invalid mode. Supported modes are 'b64_json' and 's3_json'.")
|
108 |
|
109 |
class InpaintingBatcher(AsyncBatcher):
|
110 |
-
def
|
111 |
-
super().__init__(max_batch_size)
|
112 |
-
|
113 |
-
async def process_batch(self, batch: Tuple[List[str], List[str], List[InpaintingRequest]]) -> List[Dict[str, Any]]:
|
114 |
"""
|
115 |
Process a batch of images and requests for inpainting inference.
|
116 |
|
117 |
Args:
|
118 |
-
batch (Tuple[List[str], List[
|
119 |
|
120 |
Returns:
|
121 |
List[Dict[str, Any]]: List of resulting images in the specified mode ('b64_json' or 's3_json').
|
122 |
"""
|
123 |
-
image_paths,
|
124 |
results = []
|
125 |
-
for image_path,
|
126 |
-
|
127 |
-
|
128 |
-
results.append(result)
|
129 |
-
except Exception as e:
|
130 |
-
results.append({"error": str(e)})
|
131 |
return results
|
132 |
|
133 |
@router.post("/inpainting")
|
134 |
async def inpainting_inference(
|
135 |
image: UploadFile = File(...),
|
136 |
-
mask_image: UploadFile = File(...),
|
137 |
request_data: str = Form(...),
|
138 |
):
|
139 |
"""
|
@@ -141,7 +162,6 @@ async def inpainting_inference(
|
|
141 |
|
142 |
Args:
|
143 |
image (UploadFile): Uploaded image file.
|
144 |
-
mask_image (UploadFile): Uploaded mask image file.
|
145 |
request_data (str): JSON string of the request parameters.
|
146 |
|
147 |
Returns:
|
@@ -152,10 +172,9 @@ async def inpainting_inference(
|
|
152 |
"""
|
153 |
try:
|
154 |
image_path = await save_image(image)
|
155 |
-
mask_image_path = await save_image(mask_image)
|
156 |
request_dict = json.loads(request_data)
|
157 |
request = InpaintingRequest(**request_dict)
|
158 |
-
result = run_inference(cfg, image_path,
|
159 |
return result
|
160 |
except Exception as e:
|
161 |
raise HTTPException(status_code=500, detail=str(e))
|
@@ -163,7 +182,6 @@ async def inpainting_inference(
|
|
163 |
@router.post("/inpainting/batch")
|
164 |
async def inpainting_batch_inference(
|
165 |
images: List[UploadFile] = File(...),
|
166 |
-
mask_images: List[UploadFile] = File(...),
|
167 |
request_data: str = Form(...),
|
168 |
):
|
169 |
"""
|
@@ -171,7 +189,6 @@ async def inpainting_batch_inference(
|
|
171 |
|
172 |
Args:
|
173 |
images (List[UploadFile]): List of uploaded image files.
|
174 |
-
mask_images (List[UploadFile]): List of uploaded mask image files.
|
175 |
request_data (str): JSON string of the request parameters.
|
176 |
|
177 |
Returns:
|
@@ -185,13 +202,12 @@ async def inpainting_batch_inference(
|
|
185 |
batch_request = InpaintingBatchRequestModel(**request_dict)
|
186 |
requests = batch_request.requests
|
187 |
|
188 |
-
if len(images) != len(requests)
|
189 |
-
raise HTTPException(status_code=400, detail="The number of images
|
190 |
|
191 |
batcher = InpaintingBatcher(max_batch_size=64)
|
192 |
-
image_paths = [save_image(image) for image in images]
|
193 |
-
|
194 |
-
results = await batcher.process_batch((image_paths, mask_image_paths, requests))
|
195 |
|
196 |
return results
|
197 |
except Exception as e:
|
|
|
|
|
1 |
import os
|
2 |
import uuid
|
3 |
from typing import List, Tuple, Any, Dict
|
4 |
+
from fastapi import APIRouter, File, UploadFile, HTTPException, Form
|
5 |
from pydantic import BaseModel, Field
|
6 |
from PIL import Image
|
7 |
import lightning.pytorch as pl
|
8 |
from scripts.api_utils import pil_to_s3_json, pil_to_b64_json, ImageAugmentation, accelerator
|
9 |
+
from scripts.inpainting_pipeline import AutoPaintingPipeline, load_pipeline
|
10 |
from hydra import compose, initialize
|
11 |
from async_batcher.batcher import AsyncBatcher
|
12 |
import json
|
13 |
from functools import lru_cache
|
|
|
14 |
pl.seed_everything(42)
|
15 |
router = APIRouter()
|
16 |
|
17 |
+
|
18 |
with initialize(version_base=None, config_path="../../configs"):
|
19 |
cfg = compose(config_name="inpainting")
|
20 |
|
|
|
29 |
"""
|
30 |
pipeline = load_pipeline(cfg.model, accelerator(), enable_compile=True)
|
31 |
return pipeline
|
|
|
32 |
inpainting_pipeline = load_pipeline_wrapper()
|
33 |
|
34 |
class InpaintingRequest(BaseModel):
|
|
|
42 |
guidance_scale: float = Field(..., description="Guidance scale for inference")
|
43 |
mode: str = Field(..., description="Mode for output ('b64_json' or 's3_json')")
|
44 |
num_images: int = Field(..., description="Number of images to generate")
|
45 |
+
use_augmentation: bool = Field(True, description="Whether to use image augmentation")
|
46 |
+
|
47 |
class InpaintingBatchRequestModel(BaseModel):
|
48 |
"""
|
49 |
Model representing a batch request for inpainting inference.
|
|
|
66 |
f.write(await image.read())
|
67 |
return file_path
|
68 |
|
69 |
+
def augment_image(image_path, target_width, target_height, roi_scale, segmentation_model_name, detection_model_name):
|
70 |
+
"""
|
71 |
+
Augment an image by extending its dimensions and generating masks.
|
72 |
+
|
73 |
+
Args:
|
74 |
+
image_path (str): Path to the image file.
|
75 |
+
target_width (int): Target width for augmentation.
|
76 |
+
target_height (int): Target height for augmentation.
|
77 |
+
roi_scale (float): Scale factor for region of interest.
|
78 |
+
segmentation_model_name (str): Name of the segmentation model.
|
79 |
+
detection_model_name (str): Name of the detection model.
|
80 |
+
|
81 |
+
Returns:
|
82 |
+
Tuple[Image.Image, Image.Image]: Augmented image and inverted mask.
|
83 |
+
"""
|
84 |
+
image = Image.open(image_path)
|
85 |
+
image_augmentation = ImageAugmentation(target_width, target_height, roi_scale)
|
86 |
+
image = image_augmentation.extend_image(image)
|
87 |
+
mask = image_augmentation.generate_mask_from_bbox(image, segmentation_model_name, detection_model_name)
|
88 |
+
inverted_mask = image_augmentation.invert_mask(mask)
|
89 |
+
return image, inverted_mask
|
90 |
+
|
91 |
+
def run_inference(cfg, image_path: str, request: InpaintingRequest):
|
92 |
"""
|
93 |
Run inference using an inpainting pipeline on an image.
|
94 |
|
95 |
Args:
|
96 |
cfg (dict): Configuration dictionary.
|
97 |
image_path (str): Path to the image file.
|
|
|
98 |
request (InpaintingRequest): Pydantic model containing inference parameters.
|
99 |
|
100 |
Returns:
|
|
|
103 |
Raises:
|
104 |
ValueError: If an invalid mode is provided.
|
105 |
"""
|
106 |
+
if request.use_augmentation:
|
107 |
+
image, mask_image = augment_image(image_path,
|
108 |
+
cfg['target_width'],
|
109 |
+
cfg['target_height'],
|
110 |
+
cfg['roi_scale'],
|
111 |
+
cfg['segmentation_model'],
|
112 |
+
cfg['detection_model'])
|
113 |
+
else:
|
114 |
+
image = Image.open(image_path)
|
115 |
+
mask_image = None
|
116 |
+
|
117 |
painting_pipeline = AutoPaintingPipeline(
|
118 |
pipeline=inpainting_pipeline,
|
119 |
image=image,
|
|
|
135 |
raise ValueError("Invalid mode. Supported modes are 'b64_json' and 's3_json'.")
|
136 |
|
137 |
class InpaintingBatcher(AsyncBatcher):
|
138 |
+
async def process_batch(self, batch: Tuple[List[str], List[InpaintingRequest]]) -> List[Dict[str, Any]]:
|
|
|
|
|
|
|
139 |
"""
|
140 |
Process a batch of images and requests for inpainting inference.
|
141 |
|
142 |
Args:
|
143 |
+
batch (Tuple[List[str], List[InpaintingRequest]]): Tuple of image paths and corresponding requests.
|
144 |
|
145 |
Returns:
|
146 |
List[Dict[str, Any]]: List of resulting images in the specified mode ('b64_json' or 's3_json').
|
147 |
"""
|
148 |
+
image_paths, requests = batch
|
149 |
results = []
|
150 |
+
for image_path, request in zip(image_paths, requests):
|
151 |
+
result = run_inference(cfg, image_path, request)
|
152 |
+
results.append(result)
|
|
|
|
|
|
|
153 |
return results
|
154 |
|
155 |
@router.post("/inpainting")
|
156 |
async def inpainting_inference(
|
157 |
image: UploadFile = File(...),
|
|
|
158 |
request_data: str = Form(...),
|
159 |
):
|
160 |
"""
|
|
|
162 |
|
163 |
Args:
|
164 |
image (UploadFile): Uploaded image file.
|
|
|
165 |
request_data (str): JSON string of the request parameters.
|
166 |
|
167 |
Returns:
|
|
|
172 |
"""
|
173 |
try:
|
174 |
image_path = await save_image(image)
|
|
|
175 |
request_dict = json.loads(request_data)
|
176 |
request = InpaintingRequest(**request_dict)
|
177 |
+
result = run_inference(cfg, image_path, request)
|
178 |
return result
|
179 |
except Exception as e:
|
180 |
raise HTTPException(status_code=500, detail=str(e))
|
|
|
182 |
@router.post("/inpainting/batch")
|
183 |
async def inpainting_batch_inference(
|
184 |
images: List[UploadFile] = File(...),
|
|
|
185 |
request_data: str = Form(...),
|
186 |
):
|
187 |
"""
|
|
|
189 |
|
190 |
Args:
|
191 |
images (List[UploadFile]): List of uploaded image files.
|
|
|
192 |
request_data (str): JSON string of the request parameters.
|
193 |
|
194 |
Returns:
|
|
|
202 |
batch_request = InpaintingBatchRequestModel(**request_dict)
|
203 |
requests = batch_request.requests
|
204 |
|
205 |
+
if len(images) != len(requests):
|
206 |
+
raise HTTPException(status_code=400, detail="The number of images and requests must match.")
|
207 |
|
208 |
batcher = InpaintingBatcher(max_batch_size=64)
|
209 |
+
image_paths = [await save_image(image) for image in images]
|
210 |
+
results = batcher.process_batch((image_paths, requests))
|
|
|
211 |
|
212 |
return results
|
213 |
except Exception as e:
|
api/routers/sdxl_text_to_image.py
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
import config
|
2 |
from fastapi import APIRouter, HTTPException
|
3 |
from typing import List
|
4 |
from diffusers import DiffusionPipeline
|
|
|
1 |
+
import scripts.config as config
|
2 |
from fastapi import APIRouter, HTTPException
|
3 |
from typing import List
|
4 |
from diffusers import DiffusionPipeline
|
api/yolov8l.pt.REMOVED.git-id
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
85f4cb9e9454a5944aebcc4ac73fd5571f3ddcd3
|
picpilot.egg-info/SOURCES.txt
CHANGED
@@ -16,8 +16,9 @@ scripts/config.py
|
|
16 |
scripts/inpainting_pipeline.py
|
17 |
scripts/kandinsky3_inpainting.py
|
18 |
scripts/load_pipeline.py
|
19 |
-
scripts/logger.py
|
20 |
scripts/products10k_captions.py
|
21 |
scripts/s3_manager.py
|
22 |
scripts/sdxl_lora_inference.py
|
23 |
-
scripts/sdxl_lora_tuner.py
|
|
|
|
|
|
16 |
scripts/inpainting_pipeline.py
|
17 |
scripts/kandinsky3_inpainting.py
|
18 |
scripts/load_pipeline.py
|
|
|
19 |
scripts/products10k_captions.py
|
20 |
scripts/s3_manager.py
|
21 |
scripts/sdxl_lora_inference.py
|
22 |
+
scripts/sdxl_lora_tuner.py
|
23 |
+
ui/__init__.py
|
24 |
+
ui/ui.py
|
picpilot.egg-info/top_level.txt
CHANGED
@@ -1,2 +1,3 @@
|
|
1 |
api
|
2 |
scripts
|
|
|
|
1 |
api
|
2 |
scripts
|
3 |
+
ui
|
sample_data/mask_image.png
ADDED
scripts/__pycache__/__init__.cpython-310.pyc
ADDED
Binary file (127 Bytes). View file
|
|
scripts/__pycache__/api_utils.cpython-310.pyc
ADDED
Binary file (5.71 kB). View file
|
|
scripts/__pycache__/config.cpython-310.pyc
ADDED
Binary file (2.72 kB). View file
|
|
scripts/__pycache__/inpainting_pipeline.cpython-310.pyc
ADDED
Binary file (2.64 kB). View file
|
|
scripts/__pycache__/s3_manager.cpython-310.pyc
ADDED
Binary file (2.4 kB). View file
|
|
scripts/api_utils.py
CHANGED
@@ -3,10 +3,10 @@ from ultralytics import YOLO
|
|
3 |
from transformers import SamModel, SamProcessor
|
4 |
import numpy as np
|
5 |
from PIL import Image, ImageOps
|
6 |
-
from config import SEGMENTATION_MODEL_NAME, DETECTION_MODEL_NAME
|
7 |
from diffusers.utils import load_image
|
8 |
import gc
|
9 |
-
from s3_manager import S3ManagerService
|
10 |
import io
|
11 |
from io import BytesIO
|
12 |
import base64
|
|
|
3 |
from transformers import SamModel, SamProcessor
|
4 |
import numpy as np
|
5 |
from PIL import Image, ImageOps
|
6 |
+
from scripts.config import SEGMENTATION_MODEL_NAME, DETECTION_MODEL_NAME
|
7 |
from diffusers.utils import load_image
|
8 |
import gc
|
9 |
+
from scripts.s3_manager import S3ManagerService
|
10 |
import io
|
11 |
from io import BytesIO
|
12 |
import base64
|
scripts/inpainting_pipeline.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import torch
|
2 |
from diffusers import AutoPipelineForInpainting,DiffusionPipeline
|
3 |
from diffusers.utils import load_image
|
4 |
-
from api_utils import accelerator, ImageAugmentation
|
5 |
import hydra
|
6 |
from omegaconf import DictConfig
|
7 |
from PIL import Image
|
|
|
1 |
import torch
|
2 |
from diffusers import AutoPipelineForInpainting,DiffusionPipeline
|
3 |
from diffusers.utils import load_image
|
4 |
+
from scripts.api_utils import accelerator, ImageAugmentation
|
5 |
import hydra
|
6 |
from omegaconf import DictConfig
|
7 |
from PIL import Image
|
scripts/kandinsky3_inpainting.py
CHANGED
@@ -1,12 +1,11 @@
|
|
1 |
-
import sys
|
2 |
import torch
|
3 |
-
sys.path.append('../')
|
4 |
from Kandinsky.kandinsky3 import get_inpainting_pipeline
|
5 |
-
from api_utils import ImageAugmentation
|
6 |
from diffusers.utils import load_image
|
|
|
7 |
from PIL import Image
|
8 |
|
9 |
-
device_map = torch.device(
|
10 |
dtype_map = {
|
11 |
'unet': torch.float16,
|
12 |
'text_encoder': torch.float16,
|
@@ -18,12 +17,13 @@ pipe = get_inpainting_pipeline(
|
|
18 |
device_map, dtype_map,
|
19 |
)
|
20 |
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
mask_image =
|
25 |
-
mask_image = augmenter.invert_mask(mask_image)
|
26 |
|
27 |
|
28 |
-
|
|
|
|
|
29 |
image.save('output.jpg')
|
|
|
|
|
1 |
import torch
|
|
|
2 |
from Kandinsky.kandinsky3 import get_inpainting_pipeline
|
3 |
+
from scripts.api_utils import ImageAugmentation,accelerator
|
4 |
from diffusers.utils import load_image
|
5 |
+
import numpy as np
|
6 |
from PIL import Image
|
7 |
|
8 |
+
device_map = torch.device(accelerator())
|
9 |
dtype_map = {
|
10 |
'unet': torch.float16,
|
11 |
'text_encoder': torch.float16,
|
|
|
17 |
device_map, dtype_map,
|
18 |
)
|
19 |
|
20 |
+
image = Image.open('/home/PicPilot/sample_data/image.png')
|
21 |
+
mask_image = Image.open('/home/PicPilot/sample_data/mask_image.png')
|
22 |
+
image = load_image(image=image)
|
23 |
+
mask_image = np.array(mask_image)
|
|
|
24 |
|
25 |
|
26 |
+
|
27 |
+
|
28 |
+
image = pipe( "Product on the Kitchen used for cooking", image, mask_image)
|
29 |
image.save('output.jpg')
|
scripts/load_pipeline.py
CHANGED
@@ -2,7 +2,7 @@ from config import MODEL_NAME,ADAPTER_NAME
|
|
2 |
import torch
|
3 |
from diffusers import DiffusionPipeline
|
4 |
from wandb.integration.diffusers import autolog
|
5 |
-
from config import PROJECT_NAME
|
6 |
autolog(init=dict(project=PROJECT_NAME))
|
7 |
|
8 |
|
|
|
2 |
import torch
|
3 |
from diffusers import DiffusionPipeline
|
4 |
from wandb.integration.diffusers import autolog
|
5 |
+
from scripts.config import PROJECT_NAME
|
6 |
autolog(init=dict(project=PROJECT_NAME))
|
7 |
|
8 |
|
ui/__pycache__/ui.cpython-310.pyc
ADDED
Binary file (3.55 kB). View file
|
|
ui/ui.py
CHANGED
@@ -1,92 +1,41 @@
|
|
1 |
import gradio as gr
|
2 |
import numpy as np
|
3 |
import requests
|
4 |
-
from io import BytesIO
|
5 |
import json
|
6 |
from PIL import Image
|
7 |
-
from pydantic import BaseModel, Field
|
8 |
from diffusers.utils import load_image
|
|
|
9 |
|
10 |
# API endpoints
|
11 |
-
sdxl_inference_endpoint = '
|
12 |
-
kandinsky_inpainting_inference = '
|
13 |
-
|
14 |
-
class InputRequestSDXL(BaseModel):
|
15 |
-
prompt: str
|
16 |
-
num_inference_steps: int
|
17 |
-
guidance_scale: float
|
18 |
-
negative_prompt: str
|
19 |
-
num_images: int
|
20 |
-
mode: str
|
21 |
-
|
22 |
-
class InpaintingRequestKandinsky(BaseModel):
|
23 |
-
prompt: str = Field(..., description="Prompt text for inference")
|
24 |
-
negative_prompt: str = Field(..., description="Negative prompt text for inference")
|
25 |
-
num_inference_steps: int = Field(..., description="Number of inference steps")
|
26 |
-
strength: float = Field(..., description="Strength of the inference")
|
27 |
-
guidance_scale: float = Field(..., description="Guidance scale for inference")
|
28 |
-
mode: str = Field(..., description="Mode for output ('b64_json' or 's3_json')")
|
29 |
-
num_images: int = Field(..., description="Number of images to generate")
|
30 |
|
31 |
def generate_sdxl_lora_image(prompt, negative_prompt, num_inference_steps, guidance_scale, num_images, mode):
|
32 |
-
payload =
|
33 |
-
prompt
|
34 |
-
negative_prompt
|
35 |
-
num_inference_steps
|
36 |
-
guidance_scale
|
37 |
-
num_images
|
38 |
-
mode
|
39 |
-
|
40 |
-
|
41 |
-
try:
|
42 |
-
response = requests.post(sdxl_inference_endpoint, json=payload)
|
43 |
-
response.raise_for_status()
|
44 |
-
response_data = response.json()
|
45 |
-
url = response_data['url']
|
46 |
-
image = load_image(url)
|
47 |
-
return image
|
48 |
-
except requests.exceptions.RequestException as e:
|
49 |
-
print(f"Error in SDXL-Lora API request: {e}")
|
50 |
-
return None
|
51 |
-
|
52 |
-
def process_masked_image(img):
|
53 |
-
if img is None or "composite" not in img:
|
54 |
-
return None, None
|
55 |
-
|
56 |
-
base_image = Image.fromarray(img["composite"]).convert("RGB")
|
57 |
-
|
58 |
-
if "layers" in img and len(img["layers"]) > 0:
|
59 |
-
alpha_channel = img["layers"][0][:, :, 3]
|
60 |
-
mask = np.where(alpha_channel == 0, 0, 255).astype(np.uint8)
|
61 |
-
mask = Image.fromarray(mask).convert("L")
|
62 |
-
else:
|
63 |
-
mask = Image.new("L", base_image.size, 0)
|
64 |
-
|
65 |
-
return base_image, mask
|
66 |
-
|
67 |
-
def generate_outpainting(prompt, negative_prompt, num_inference_steps, strength, guidance_scale, mode, num_images, masked_image, width, height):
|
68 |
-
base_image, mask = process_masked_image(masked_image)
|
69 |
-
|
70 |
-
if base_image is None or mask is None:
|
71 |
-
return None, None
|
72 |
-
|
73 |
-
# Resize base image and mask
|
74 |
-
base_image_resized = base_image.resize((width, height))
|
75 |
-
mask_resized = mask.resize((width, height))
|
76 |
|
77 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
78 |
img_byte_arr = BytesIO()
|
79 |
-
|
80 |
img_byte_arr = img_byte_arr.getvalue()
|
81 |
|
82 |
-
mask_byte_arr = BytesIO()
|
83 |
-
mask_resized.save(mask_byte_arr, format='PNG')
|
84 |
-
mask_byte_arr = mask_byte_arr.getvalue()
|
85 |
-
|
86 |
# Prepare the files for multipart/form-data
|
87 |
files = {
|
88 |
-
'image': ('image.png', img_byte_arr, 'image/png')
|
89 |
-
'mask_image': ('mask.png', mask_byte_arr, 'image/png'),
|
90 |
}
|
91 |
|
92 |
# Prepare the request data
|
@@ -97,7 +46,9 @@ def generate_outpainting(prompt, negative_prompt, num_inference_steps, strength,
|
|
97 |
"strength": strength,
|
98 |
"guidance_scale": guidance_scale,
|
99 |
"mode": mode,
|
100 |
-
"num_images": num_images
|
|
|
|
|
101 |
}
|
102 |
|
103 |
# Convert request_data to a JSON string
|
@@ -108,30 +59,15 @@ def generate_outpainting(prompt, negative_prompt, num_inference_steps, strength,
|
|
108 |
'request_data': request_data_json
|
109 |
}
|
110 |
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
print(f"Error in Kandinsky Inpainting API request: {e}")
|
120 |
-
return None, None
|
121 |
-
|
122 |
-
def generate_mask_preview(img):
|
123 |
-
base_image, mask = process_masked_image(img)
|
124 |
-
if mask is None:
|
125 |
-
return None
|
126 |
-
return mask
|
127 |
-
|
128 |
-
def resize_image(img, width, height):
|
129 |
-
if img is None:
|
130 |
-
return None
|
131 |
-
resized_img = img.resize((width, height))
|
132 |
-
return resized_img
|
133 |
|
134 |
-
# Gradio interface setup
|
135 |
with gr.Blocks(theme='VikramSingh178/Webui-Theme') as demo:
|
136 |
with gr.Tab("SdxL-Lora"):
|
137 |
with gr.Row():
|
@@ -149,32 +85,25 @@ with gr.Blocks(theme='VikramSingh178/Webui-Theme') as demo:
|
|
149 |
image_preview = gr.Image(label="Generated Image (SDXL-Lora)", show_download_button=True, show_share_button=True, container=True)
|
150 |
generate_button.click(generate_sdxl_lora_image, inputs=[prompt, negative_prompt, num_inference_steps, guidance_scale, num_images, mode], outputs=[image_preview])
|
151 |
|
152 |
-
|
153 |
with gr.Tab("Inpainting"):
|
154 |
with gr.Row():
|
155 |
with gr.Column():
|
156 |
with gr.Group():
|
157 |
-
|
158 |
prompt = gr.Textbox(label="Prompt", placeholder="Enter your prompt here")
|
159 |
negative_prompt= gr.Textbox(label="Negative Prompt", placeholder="Enter negative prompt here")
|
160 |
num_inference_steps = gr.Slider(minimum=1, maximum=100, step=1, value=20, label="Inference Steps")
|
161 |
strength = gr.Slider(minimum=0.1, maximum=1, step=0.1, value=0.8, label="Strength")
|
162 |
guidance_scale = gr.Slider(minimum=1.0, maximum=10.0, step=0.1, value=7.5, label="Guidance Scale")
|
163 |
-
num_images= gr.Slider(minimum=1, maximum=10, step=1, value=1, label="Number of Images")
|
164 |
mode_kandinsky = gr.Dropdown(choices=["s3_json", "b64_json"], value="s3_json", label="Mode")
|
165 |
width_slider = gr.Slider(minimum=512, maximum=1024, step=1, value=800, label="Image Width")
|
166 |
height_slider = gr.Slider(minimum=512, maximum=1024, step=1, value=800, label="Image Height")
|
167 |
-
resize_button = gr.Button("Resize Image", variant='secondary')
|
168 |
generate_button = gr.Button("Generate Inpainting", variant='primary')
|
169 |
-
generate_mask_button_painting = gr.Button("Generate Mask", variant='primary')
|
170 |
|
171 |
with gr.Column(scale=1):
|
172 |
-
mask_preview= gr.Image(label="Mask Preview", show_download_button=True, container=True)
|
173 |
outpainted_image_preview = gr.Image(label="Outpainted Image (Kandinsky)", show_download_button=True, show_share_button=True, container=True)
|
174 |
-
|
175 |
-
|
176 |
-
generate_button.click(generate_outpainting,
|
177 |
-
inputs=[prompt, negative_prompt, num_inference_steps, strength, guidance_scale, mode_kandinsky, num_images, masked_image, width_slider, height_slider],
|
178 |
-
outputs=[mask_preview, outpainted_image_preview])
|
179 |
|
180 |
-
demo.launch()
|
|
|
1 |
import gradio as gr
|
2 |
import numpy as np
|
3 |
import requests
|
|
|
4 |
import json
|
5 |
from PIL import Image
|
|
|
6 |
from diffusers.utils import load_image
|
7 |
+
from io import BytesIO
|
8 |
|
9 |
# API endpoints
|
10 |
+
sdxl_inference_endpoint = 'http://localhost:7860/api/v1/product-diffusion/sdxl_v0_lora_inference'
|
11 |
+
kandinsky_inpainting_inference = 'http://localhost:7860/api/v1/product-diffusion/inpainting'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
|
13 |
def generate_sdxl_lora_image(prompt, negative_prompt, num_inference_steps, guidance_scale, num_images, mode):
|
14 |
+
payload = {
|
15 |
+
"prompt": prompt,
|
16 |
+
"negative_prompt": negative_prompt,
|
17 |
+
"num_inference_steps": num_inference_steps,
|
18 |
+
"guidance_scale": guidance_scale,
|
19 |
+
"num_images": num_images,
|
20 |
+
"mode": mode
|
21 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
+
response = requests.post(sdxl_inference_endpoint, json=payload)
|
24 |
+
response.raise_for_status()
|
25 |
+
response_data = response.json()
|
26 |
+
url = response_data['url']
|
27 |
+
image = load_image(url)
|
28 |
+
return image
|
29 |
+
|
30 |
+
def generate_outpainting(prompt, negative_prompt, num_inference_steps, strength, guidance_scale, mode, num_images, image, width, height):
|
31 |
+
# Convert the image to bytes
|
32 |
img_byte_arr = BytesIO()
|
33 |
+
image.save(img_byte_arr, format='PNG')
|
34 |
img_byte_arr = img_byte_arr.getvalue()
|
35 |
|
|
|
|
|
|
|
|
|
36 |
# Prepare the files for multipart/form-data
|
37 |
files = {
|
38 |
+
'image': ('image.png', img_byte_arr, 'image/png')
|
|
|
39 |
}
|
40 |
|
41 |
# Prepare the request data
|
|
|
46 |
"strength": strength,
|
47 |
"guidance_scale": guidance_scale,
|
48 |
"mode": mode,
|
49 |
+
"num_images": num_images,
|
50 |
+
"width": width,
|
51 |
+
"height": height
|
52 |
}
|
53 |
|
54 |
# Convert request_data to a JSON string
|
|
|
59 |
'request_data': request_data_json
|
60 |
}
|
61 |
|
62 |
+
response = requests.post(kandinsky_inpainting_inference, files=files, data=form_data)
|
63 |
+
response.raise_for_status()
|
64 |
+
response_data = response.json()
|
65 |
+
image_url = response_data['image_url']
|
66 |
+
mask_url = response_data['mask_url']
|
67 |
+
outpainted_image = load_image(image_url)
|
68 |
+
mask_image = load_image(mask_url)
|
69 |
+
return outpainted_image, mask_image
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
70 |
|
|
|
71 |
with gr.Blocks(theme='VikramSingh178/Webui-Theme') as demo:
|
72 |
with gr.Tab("SdxL-Lora"):
|
73 |
with gr.Row():
|
|
|
85 |
image_preview = gr.Image(label="Generated Image (SDXL-Lora)", show_download_button=True, show_share_button=True, container=True)
|
86 |
generate_button.click(generate_sdxl_lora_image, inputs=[prompt, negative_prompt, num_inference_steps, guidance_scale, num_images, mode], outputs=[image_preview])
|
87 |
|
|
|
88 |
with gr.Tab("Inpainting"):
|
89 |
with gr.Row():
|
90 |
with gr.Column():
|
91 |
with gr.Group():
|
92 |
+
input_image = gr.Image(label="Upload Image", type="pil")
|
93 |
prompt = gr.Textbox(label="Prompt", placeholder="Enter your prompt here")
|
94 |
negative_prompt= gr.Textbox(label="Negative Prompt", placeholder="Enter negative prompt here")
|
95 |
num_inference_steps = gr.Slider(minimum=1, maximum=100, step=1, value=20, label="Inference Steps")
|
96 |
strength = gr.Slider(minimum=0.1, maximum=1, step=0.1, value=0.8, label="Strength")
|
97 |
guidance_scale = gr.Slider(minimum=1.0, maximum=10.0, step=0.1, value=7.5, label="Guidance Scale")
|
98 |
+
num_images = gr.Slider(minimum=1, maximum=10, step=1, value=1, label="Number of Images")
|
99 |
mode_kandinsky = gr.Dropdown(choices=["s3_json", "b64_json"], value="s3_json", label="Mode")
|
100 |
width_slider = gr.Slider(minimum=512, maximum=1024, step=1, value=800, label="Image Width")
|
101 |
height_slider = gr.Slider(minimum=512, maximum=1024, step=1, value=800, label="Image Height")
|
|
|
102 |
generate_button = gr.Button("Generate Inpainting", variant='primary')
|
|
|
103 |
|
104 |
with gr.Column(scale=1):
|
|
|
105 |
outpainted_image_preview = gr.Image(label="Outpainted Image (Kandinsky)", show_download_button=True, show_share_button=True, container=True)
|
106 |
+
mask_image_preview = gr.Image(label="Generated Mask", show_download_button=True, show_share_button=True, container=True)
|
107 |
+
generate_button.click(generate_outpainting, inputs=[prompt, negative_prompt, num_inference_steps, strength, guidance_scale, mode_kandinsky, num_images, input_image, width_slider, height_slider], outputs=[outpainted_image_preview, mask_image_preview])
|
|
|
|
|
|
|
108 |
|
109 |
+
demo.launch()
|