-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
531 lines (435 loc) · 16.3 KB
/
app.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
from flask import Flask, request
from flask_cors import CORS
from datetime import datetime
import base64, ast, torch, cv2
from transformers import Blip2Processor, Blip2ForConditionalGeneration
from gpt_prompts import (
caption_to_layout,
keywords_to_descriptions,
keywords_expansion,
caption_to_keywords,
caption_layout_matcher,
)
from segment_anything import SamAutomaticMaskGenerator, sam_model_registry
from diffusers import StableDiffusionGLIGENPipeline
from style_module.style_transfer import line_drawing_predict
from layout_module.layout_metrics import generate_layouts, xywh_to_xyxy
DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# blipprocessor = None
# blipmodel = None
# sam = None
# mask_generator = None
# Setup BLIP model
blipprocessor = Blip2Processor.from_pretrained(
"Salesforce/blip2-opt-2.7b", low_cpu_mem_usage=True
)
blipmodel = Blip2ForConditionalGeneration.from_pretrained(
"Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16, low_cpu_mem_usage=True
).to(DEVICE)
gligen = StableDiffusionGLIGENPipeline.from_pretrained(
"masterful/gligen-1-4-generation-text-box",
variant="fp16",
torch_dtype=torch.float16,
).to(DEVICE)
# Setup SAM model
MODEL_TYPE = "vit_l"
CHECKPOINT_PATH = "./checkpoints/sam_vit_l_0b3195.pth"
sam = sam_model_registry[MODEL_TYPE](checkpoint=CHECKPOINT_PATH)
sam.to(DEVICE)
mask_generator = SamAutomaticMaskGenerator(sam)
app = Flask(__name__, static_folder="./generated")
CORS(app, resources={r"/*": {"origins": "*"}})
from log import create_log_api
app.register_blueprint(create_log_api(), url_prefix="/log")
# Util functions
def keywordlist_to_string(keywords):
matters = []
actions = []
themes = []
for e in keywords:
if e["type"] == "Subject matter":
matters.append(e)
elif e["type"] == "Action & pose":
actions.append(e)
elif e["type"] == "Theme & mood":
themes.append(e)
res = ""
if len(matters) > 0:
res += (
"Subject matter: "
+ ", ".join([o["keyword"].split("-")[0] for o in matters])
+ "\n"
)
if len(actions) > 0:
res += "Action & pose: " + ", ".join([a["keyword"] for a in actions]) + "\n"
if len(themes) > 0:
res += "Theme & mood: " + ", ".join([c["keyword"] for c in themes])
res = res.strip()
return res
def prompt_to_recombined_images(input_prompt, gen_num=1):
'''
Input: prompt
e.g. """Caption: a waterfall and a modern high speed train running through the tunnel in a beautiful forest with fall foliage.
Objects: [('a waterfall', [0.1387, 0.2051, 0.4277, 0.7090]), ('a modern high speed train running through the tunnel', [0.4980, 0.4355, 0.8516, 0.7266])]
"""
% Now: input bbox is xywh format, ratio. <-- It can be change!!
Output:
- image_path_raw: generated raw image path list
- image_path_sketch: generated sketch image path list
'''
timestamp = datetime.now().strftime("%Y%m%d%H%M%S")
image_path_sketch_list = [
"generated/" + f"{timestamp}_{i}_sketch.png" for i in range(gen_num)
]
image_path_raw_list = [
"generated/" + f"{timestamp}_{i}_raw.png" for i in range(gen_num)
]
def parse_input(text=None):
try:
if "Objects: " in text:
caption, objects = text.split("Objects: ")
caption = caption.replace("Caption: ", "")
objects = ast.literal_eval(objects)
except Exception as e:
raise Exception(f"response format invalid: {e} (text: {text})")
return caption, objects
input_prompt = parse_input(input_prompt)
prompt = input_prompt[0]
boxes = xywh_to_xyxy(
[i[1] for i in input_prompt[1]]
) # [[0.1387, 0.2051, 0.4277, 0.7090], [0.4980, 0.4355, 0.8516, 0.7266]]
phrases = [
i[0] for i in input_prompt[1]
] # ["a waterfall", "a modern high speed train running through the tunnel"]
images = gligen(
prompt=prompt,
gligen_phrases=phrases,
gligen_boxes=boxes,
gligen_scheduled_sampling_beta=1,
output_type="pil",
num_inference_steps=15,
num_images_per_prompt=gen_num,
).images
for image, image_path_raw, image_path_sketch in zip(
images, image_path_raw_list, image_path_sketch_list
):
image.save(image_path_raw)
output_sketch = line_drawing_predict(image, ver="Simple Lines")
output_sketch.save(image_path_sketch)
return image_path_raw_list, image_path_sketch_list
# Routes
@app.route("/setup", methods=["GET"])
def setup():
# global blipprocessor, blipmodel, sam, mask_generator
# # Setup BLIP model
# blipprocessor = Blip2Processor.from_pretrained(
# "Salesforce/blip2-opt-2.7b", low_cpu_mem_usage=True
# )
# blipmodel = Blip2ForConditionalGeneration.from_pretrained(
# "Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16, low_cpu_mem_usage=True
# ).to(DEVICE)
# # Setup SAM model
# MODEL_TYPE = "vit_l"
# CHECKPOINT_PATH = "./checkpoints/sam_vit_l_0b3195.pth"
# sam = sam_model_registry[MODEL_TYPE](checkpoint=CHECKPOINT_PATH)
# sam.to(DEVICE)
# mask_generator = SamAutomaticMaskGenerator(sam)
return "setup done"
@app.route("/", methods=["GET"])
def test():
return "hello, world!"
@app.route("/sendImage", methods=["POST"])
def store_image():
"""
Input:
- image: base64 encoded image
Output:
- filename: filename of uploaded image
"""
data = request.get_json()
image = data.get("image")
if not image:
return {"message": "No file received"}, 400
# Save image
file_content = image.split(";base64,")[1]
file_extension = image.split(";")[0].split("/")[1]
timestamp = datetime.now().strftime("%Y%m%d%H%M%S")
filename = f"{timestamp}.{file_extension}"
with open("uploaded/" + filename, "wb") as f:
f.write(base64.b64decode(file_content))
return {"filename": filename}, 200
# Image --> Keyword list
@app.route("/imageToKeywords", methods=["POST"])
def extract_element_from_image():
"""
Input:
- image: base64 encoded image
Output:
- filename: filename of uploaded image
- keywords: list of keywords
"""
data = request.get_json()
image = data.get("image")
if not image:
return {"message": "No file received"}, 400
# Save image
file_content = image.split(";base64,")[1]
file_extension = image.split(";")[0].split("/")[1]
timestamp = datetime.now().strftime("%Y%m%d%H%M%S")
filename = f"{timestamp}.{file_extension}"
with open("uploaded/" + filename, "wb") as f:
f.write(base64.b64decode(file_content))
# Load image
IMAGE_PATH = "uploaded/" + filename
image_bgr = cv2.imread(IMAGE_PATH)
image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
# Get whole image description
segmented_descriptions = []
inputs = blipprocessor(images=image_rgb, return_tensors="pt").to(
DEVICE, torch.float16
)
generated_ids = blipmodel.generate(**inputs)
whole_image_description = blipprocessor.batch_decode(
generated_ids, skip_special_tokens=True
)[0].strip()
segmented_descriptions.append(whole_image_description)
# Get segmented image descriptions
image_segments = []
for i in range(3):
for j in range(3):
w = int(image_rgb.shape[0] / 2)
h = int(image_rgb.shape[1] / 2)
x = int(image_rgb.shape[0] / 4 * i)
y = int(image_rgb.shape[1] / 4 * j)
image_segments.append(image_rgb[x : x + w, y : y + h])
for seg in image_segments:
inputs = blipprocessor(images=seg, return_tensors="pt").to(
DEVICE, torch.float16
)
generated_ids = blipmodel.generate(**inputs)
generated_text = blipprocessor.batch_decode(
generated_ids, skip_special_tokens=True
)[0].strip()
segmented_descriptions.append(generated_text)
segmented_descriptions = "\n".join(list(set(segmented_descriptions)))
matters, _, _ = caption_to_keywords(whole_image_description)
_, actions, themes = caption_to_keywords(segmented_descriptions)
keywords = []
for matter in matters:
keywords.append({"type": "Subject matter", "keyword": matter, "position": None})
for action in actions:
keywords.append({"type": "Action & pose", "keyword": action, "position": None})
for theme in themes:
keywords.append({"type": "Theme & mood", "keyword": theme, "position": None})
return {"filename": filename, "keywords": keywords}, 200
# Image --> Layout
@app.route("/imageToLayout", methods=["POST"])
def extract_layout_from_image():
"""
Input:
- filename: filename of uploaded image
Output:
- bboxes: list of bounding boxes (proportion of image size)
"""
data = request.get_json()
filename = data.get("filename")
if not filename:
return {"message": "No file received"}, 400
# Load image
print("Loading image")
IMAGE_PATH = "uploaded/" + filename
image_bgr = cv2.imread(IMAGE_PATH)
image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
width, height = image_rgb.shape[1], image_rgb.shape[0]
masks = mask_generator.generate(image_rgb)
# return only bbox
bboxes = []
max_mask_size = max([m["area"] for m in masks])
# sorted_masks = sorted(masks, key=(lambda x: x['predicted_iou']), reverse=True)
sorted_masks = sorted(
masks,
key=(
lambda x: x["area"]
- (x["stability_score"] < 0.2) * max_mask_size
- (x["predicted_iou"] < 0.2) * max_mask_size
),
reverse=True,
)
i = 0
while len(bboxes) < 5 and i < len(sorted_masks):
mask = sorted_masks[i]
if mask["area"] < 0.7 * width * height:
x, y, w, h = mask["bbox"]
bboxes.append((x / width, y / height, w / width, h / height))
i += 1
return {"bboxes": bboxes}, 200
# Layout --> Layout
@app.route("/getRecommendedLayouts", methods=["POST"])
def recommend_layouts():
"""
Input:
- layout: list of xywh format bounding boxes (512*512)
Output:
- layouts: 2-dim list of 10 recommended layouts (512*512)
"""
data = request.get_json()
old_layout = data.get("layout")
layouts = []
# for i in range(1, len(old_layout)+1):
# layouts.append(generate_layouts(old_layout, recommends_num=10, target_bbox_num=i))
target_bbox_num = data.get("target_bbox_num")
layouts = generate_layouts(
old_layout, recommends_num=10, target_bbox_num=target_bbox_num
)
return {"layouts": layouts}, 200
# Layout diffusion
@app.route("/generateImage", methods=["POST"])
def generate_recombined_images():
'''
Input:
- prompt: prompt for layout diffusion
e.g., """Caption: Gray cat and a soccer ball on the grass, line drawing.
Objects: [('a gray cat', [67, 243, 120, 126]), ('a soccer ball', [265, 193, 190, 210])]
"""
- gen_num: number of images from single prompt
Output:
- image_path_raw: generated raw image path list
- image_path_sketch: generated sketch image path list
'''
data = request.get_json()
prompt = data.get("prompt")
generation_image_num = data.get("gen_num")
image_path_raw_list, image_path_sketch_list = prompt_to_recombined_images(
prompt, gen_num=generation_image_num
)
return {
"image_path_raw": image_path_raw_list,
"image_path_sketch": image_path_sketch_list,
}, 200
# Keyword list --> Expanded keyword list
@app.route("/expandElements", methods=["POST"])
def generate_elements_from_elements():
"""
Input:
- originalKeywords: list of original elements
Output:
- suggestedKeywords: list of expanded elements
"""
data = request.get_json()
originalKeywords = data.get("originalKeywords")
keywordString = keywordlist_to_string(originalKeywords)
suggestedKeywords = keywords_expansion(keywordString)
return {"suggestedKeywords": suggestedKeywords}, 200
# Keyword list --> Descriptions only
@app.route("/mergeKeywordsToDescriptions", methods=["POST"])
def merge_elements_to_desc():
data = request.get_json()
keywords = data.get("keywords")
keywordString = keywordlist_to_string(keywords)
generatedDescriptions = keywords_to_descriptions(keywordString)
return {"descriptions": generatedDescriptions}, 200
# Descriptions --> Layouts & Sketches
@app.route("/descriptionToSketch", methods=["POST"])
def descriptionToSketch():
data = request.get_json()
description = data.get("description")
objects = list(description["objects"].keys())
print(objects)
layout = caption_to_layout(description["caption"], objects)
print(layout)
prompt = (
"Caption: "
+ description["caption"]
+ ",line drawing illustration"
+ "\nObjects: "
+ str(layout)
)
print(prompt)
image_path_raw, image_path_sketch = prompt_to_recombined_images(prompt, gen_num=1)
description["layout"] = ast.literal_eval(layout)
description["image_path_raw"] = image_path_raw
description["image_path_sketch"] = image_path_sketch
return {"image_path_sketch": image_path_sketch}, 200
# Keyword list --> Descriptions & Layouts & Sketches
@app.route("/mergeKeywords", methods=["POST"])
def merge_elements():
data = request.get_json()
keywords = data.get("keywords")
keywordString = keywordlist_to_string(keywords)
generatedDescriptions = keywords_to_descriptions(keywordString)
for description in generatedDescriptions:
objects = list(description["objects"].keys())
print(objects)
layout = caption_to_layout(description["caption"], objects)
print(layout)
prompt = (
"Caption: "
+ description["caption"]
+ ",line drawing illustration"
+ "\nObjects: "
+ str(layout)
)
print(prompt)
image_path_raw, image_path_sketch = prompt_to_recombined_images(prompt)
description["layout"] = ast.literal_eval(layout)
description["image_path_raw"] = image_path_raw
description["image_path_sketch"] = image_path_sketch
return {"descriptions": generatedDescriptions}, 200
@app.route("/getMoreSketches", methods=["POST"])
def generate_more_sketches():
data = request.get_json()
description = data.get("description")
old_layout = data.get("layout")
target_bbox_num = len(description["objects"])
# Generate layouts
new_layouts = []
if old_layout is None or len(old_layout) < len(description["objects"]):
original_layout = ast.literal_eval(
caption_to_layout(description["caption"], description["objects"])
)
adjusted_layout = []
for _, bbox in original_layout:
adjusted_layout.append(
[bbox[0] * 512, bbox[1] * 512, bbox[2] * 512, bbox[3] * 512]
)
layouts = generate_layouts(
adjusted_layout, recommends_num=5, target_bbox_num=target_bbox_num
)
else:
layouts = generate_layouts(
old_layout, recommends_num=5, target_bbox_num=target_bbox_num
)
for layout in layouts:
adjusted_layout = []
for bbox in layout:
adjusted_layout.append(
[bbox[0] / 512, bbox[1] / 512, bbox[2] / 512, bbox[3] / 512]
)
new_layout = caption_layout_matcher(
description["caption"], description["objects"], str(adjusted_layout)
)
new_layouts.append(new_layout)
print("new_layouts")
print(new_layouts)
image_path_raw_list = []
image_path_sketch_list = []
# Generate skteches
for new_layout in new_layouts:
prompt = (
"Caption: "
+ description["caption"]
+ ",line drawing illustration"
+ "\nObjects: "
+ str(new_layout)
)
image_path_raw, image_path_sketch = prompt_to_recombined_images(
prompt, gen_num=1
)
image_path_raw_list.append(image_path_raw[0])
image_path_sketch_list.append(image_path_sketch[0])
return {
"image_path_raw": image_path_raw_list,
"image_path_sketch": image_path_sketch_list,
}, 200
if __name__ == "__main__":
app.run(host="0.0.0.0", port=7887, debug=False)