-
Notifications
You must be signed in to change notification settings - Fork 36
/
Copy pathtext_to_360panorama_image_pipeline.py
217 lines (192 loc) · 8.34 KB
/
text_to_360panorama_image_pipeline.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
# Copyright © Alibaba, Inc. and its affiliates.
import random
from typing import Any, Dict
import numpy as np
import torch
from basicsr.archs.rrdbnet_arch import RRDBNet
from diffusers import (ControlNetModel, DiffusionPipeline,
EulerAncestralDiscreteScheduler,
UniPCMultistepScheduler)
from PIL import Image
from realesrgan import RealESRGANer
from .pipeline_base import StableDiffusionBlendExtendPipeline
from .pipeline_sr import StableDiffusionControlNetImg2ImgPanoPipeline
class Text2360PanoramaImagePipeline(DiffusionPipeline):
""" Stable Diffusion for 360 Panorama Image Generation Pipeline.
Example:
>>> import torch
>>> from txt2panoimg import Text2360PanoramaImagePipeline
>>> prompt = 'The mountains'
>>> input = {'prompt': prompt, 'upscale': True}
>>> model_id = 'models/'
>>> txt2panoimg = Text2360PanoramaImagePipeline(model_id, torch_dtype=torch.float16)
>>> output = txt2panoimg(input)
>>> output.save('result.png')
"""
def __init__(self, model: str, device: str = 'cuda', **kwargs):
"""
Use `model` to create a stable diffusion pipeline for 360 panorama image generation.
Args:
model: model id on modelscope hub.
device: str = 'cuda'
"""
super().__init__()
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu'
) if device is None else device
if device == 'gpu':
device = 'cuda'
torch_dtype = kwargs.get('torch_dtype', torch.float16)
enable_xformers_memory_efficient_attention = kwargs.get(
'enable_xformers_memory_efficient_attention', True)
model_id = model + '/sd-base/'
# init base model
self.pipe = StableDiffusionBlendExtendPipeline.from_pretrained(
model_id, torch_dtype=torch_dtype).to(device)
self.pipe.vae.enable_tiling()
self.pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(
self.pipe.scheduler.config)
# remove following line if xformers is not installed
try:
if enable_xformers_memory_efficient_attention:
self.pipe.enable_xformers_memory_efficient_attention()
except Exception as e:
print(e)
self.pipe.enable_model_cpu_offload()
# init controlnet-sr model
base_model_path = model + '/sr-base'
controlnet_path = model + '/sr-control'
controlnet = ControlNetModel.from_pretrained(
controlnet_path, torch_dtype=torch_dtype)
self.pipe_sr = StableDiffusionControlNetImg2ImgPanoPipeline.from_pretrained(
base_model_path, controlnet=controlnet,
torch_dtype=torch_dtype).to(device)
self.pipe_sr.scheduler = UniPCMultistepScheduler.from_config(
self.pipe.scheduler.config)
self.pipe_sr.vae.enable_tiling()
# remove following line if xformers is not installed
try:
if enable_xformers_memory_efficient_attention:
self.pipe_sr.enable_xformers_memory_efficient_attention()
except Exception as e:
print(e)
self.pipe_sr.enable_model_cpu_offload()
# init realesrgan model
sr_model = RRDBNet(
num_in_ch=3,
num_out_ch=3,
num_feat=64,
num_block=23,
num_grow_ch=32,
scale=2)
netscale = 2
model_path = model + '/RealESRGAN_x2plus.pth'
dni_weight = None
self.upsampler = RealESRGANer(
scale=netscale,
model_path=model_path,
dni_weight=dni_weight,
model=sr_model,
tile=384,
tile_pad=20,
pre_pad=20,
half=False,
device=device,
)
@staticmethod
def blend_h(a, b, blend_extent):
blend_extent = min(a.shape[1], b.shape[1], blend_extent)
for x in range(blend_extent):
b[:, x, :] = a[:, -blend_extent
+ x, :] * (1 - x / blend_extent) + b[:, x, :] * (
x / blend_extent)
return b
def __call__(self, inputs: Dict[str, Any],
**forward_params) -> Dict[str, Any]:
if not isinstance(inputs, dict):
raise ValueError(
f'Expected the input to be a dictionary, but got {type(input)}'
)
num_inference_steps = inputs.get('num_inference_steps', 20)
guidance_scale = inputs.get('guidance_scale', 7.5)
preset_a_prompt = 'photorealistic, trend on artstation, ((best quality)), ((ultra high res))'
add_prompt = inputs.get('add_prompt', preset_a_prompt)
preset_n_prompt = 'persons, complex texture, small objects, sheltered, blur, worst quality, '\
'low quality, zombie, logo, text, watermark, username, monochrome, '\
'complex lighting'
negative_prompt = inputs.get('negative_prompt', preset_n_prompt)
seed = inputs.get('seed', -1)
upscale = inputs.get('upscale', True)
refinement = inputs.get('refinement', True)
guidance_scale_sr_step1 = inputs.get('guidance_scale_sr_step1', 15)
guidance_scale_sr_step2 = inputs.get('guidance_scale_sr_step1', 17)
if 'prompt' in inputs.keys():
prompt = inputs['prompt']
else:
# for demo_service
prompt = forward_params.get('prompt', 'the living room')
print(f'Test with prompt: {prompt}')
if seed == -1:
seed = random.randint(0, 65535)
print(f'global seed: {seed}')
generator = torch.manual_seed(seed)
prompt = '<360panorama>, ' + prompt + ', ' + add_prompt
output_img = self.pipe(
prompt,
negative_prompt=negative_prompt,
num_inference_steps=num_inference_steps,
height=512,
width=1024,
guidance_scale=guidance_scale,
generator=generator).images[0]
if not upscale:
print('finished')
else:
print('inputs: upscale=True, running upscaler.')
print('running upscaler step1. Initial super-resolution')
sr_scale = 2.0
output_img = self.pipe_sr(
prompt.replace('<360panorama>, ', ''),
negative_prompt=negative_prompt,
image=output_img.resize(
(int(1536 * sr_scale), int(768 * sr_scale))),
num_inference_steps=7,
generator=generator,
control_image=output_img.resize(
(int(1536 * sr_scale), int(768 * sr_scale))),
strength=0.8,
controlnet_conditioning_scale=1.0,
guidance_scale=guidance_scale_sr_step1,
).images[0]
print('running upscaler step2. Super-resolution with Real-ESRGAN')
output_img = output_img.resize((1536 * 2, 768 * 2))
w = output_img.size[0]
blend_extend = 10
outscale = 2
output_img = np.array(output_img)
output_img = np.concatenate(
[output_img, output_img[:, :blend_extend, :]], axis=1)
output_img, _ = self.upsampler.enhance(
output_img, outscale=outscale)
output_img = self.blend_h(output_img, output_img,
blend_extend * outscale)
output_img = Image.fromarray(output_img[:, :w * outscale, :])
if refinement:
print(
'inputs: refinement=True, running refinement. This is a bit time-consuming.'
)
sr_scale = 4
output_img = self.pipe_sr(
prompt.replace('<360panorama>, ', ''),
negative_prompt=negative_prompt,
image=output_img.resize(
(int(1536 * sr_scale), int(768 * sr_scale))),
num_inference_steps=7,
generator=generator,
control_image=output_img.resize(
(int(1536 * sr_scale), int(768 * sr_scale))),
strength=0.8,
controlnet_conditioning_scale=1.0,
guidance_scale=guidance_scale_sr_step2,
).images[0]
print('finished')
return output_img