File size: 6,854 Bytes
c4e6a63
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations


import gradio as gr
import numpy as np
from PIL import Image

import nltk
nltk.download('punkt')
nltk.download('averaged_perceptron_tagger')

from main import LPMConfig, main

DESCRIPTION = '''# Localizing Object-level Shape Variations with Text-to-Image Diffusion Models
This is a demo for our ''Localizing Object-level Shape Variations with Text-to-Image Diffusion Models'' [paper](https://arxiv.org/abs/2303.11306).
We introduce a method that generates object-level shape variation for a given image. 
This demo allows using a real image as well as a generated image. For a real image, a matching prompt is required.
'''

def main_pipeline(
        prompt: str,
        object_of_interest: str,
        proxy_words: str,
        number_of_variations: int,
        start_prompt_range: int,
        end_prompt_range: int,
        objects_to_preserve: str,
        background_nouns: str,
        seed: int,
        input_image: str):
        prompt = prompt.replace(object_of_interest, '{word}')
        print(number_of_variations)
        print(proxy_words)
        proxy_words = proxy_words.split(',') if proxy_words != '' else []
        objects_to_preserve = objects_to_preserve.split(',') if objects_to_preserve != '' else []
        background_nouns = background_nouns.split(',') if background_nouns != '' else []
        args = LPMConfig(
            seed=seed,
            prompt=prompt,
            object_of_interest=object_of_interest,
            proxy_words=proxy_words,
            number_of_variations=number_of_variations,
            start_prompt_range=start_prompt_range,
            end_prompt_range=end_prompt_range,
            objects_to_preserve=objects_to_preserve,
            background_nouns=background_nouns,
            real_image_path="" if input_image is None else input_image
        )

        result_images, result_proxy_words = main(args)
        result_images = [im.permute(1, 2, 0).cpu().numpy() for im in result_images]
        result_images = [(im * 255).astype(np.uint8) for im in result_images]
        result_images = [Image.fromarray(im) for im in result_images]

        return result_images, ",".join(result_proxy_words)


with gr.Blocks(css='style.css') as demo:
    gr.Markdown(DESCRIPTION)

    with gr.Row():
        with gr.Column():
            input_image = gr.Image(
                label="Input image (optional)",
                type="filepath"
            )
            prompt = gr.Text(
                label='Prompt',
                max_lines=1,
                placeholder='A table below a lamp',
            )
            object_of_interest = gr.Text(
                label='Object of interest',
                max_lines=1,
                placeholder='lamp',
            )
            proxy_words = gr.Text(
                label='Proxy words - words used to obtain variations (a comma-separated list of words, can leave empty)',
                max_lines=1,
                placeholder=''
            )
            number_of_variations = gr.Slider(
                label='Number of variations (used only for automatic proxy-words)',
                minimum=2,
                maximum=30,
                value=20,
                step=1
            )
            start_prompt_range = gr.Slider(
                label='Number of steps before starting shape interval',
                minimum=0,
                maximum=50,
                value=7,
                step=1
            )
            end_prompt_range = gr.Slider(
                label='Number of steps before ending shape interval',
                minimum=1,
                maximum=50,
                value=17,
                step=1
            )
            objects_to_preserve = gr.Text(
                label='Words corresponding to objects to preserve (a comma-separated list of words, can leave empty)',
                max_lines=1,
                placeholder='table',
            )
            background_nouns = gr.Text(
                label='Words corresponding to objects that should be copied from original image (a comma-separated list of words, can leave empty)',
                max_lines=1,
                placeholder='',
            )
            seed = gr.Slider(
                label='Seed',
                minimum=1,
                maximum=100000,
                value=0,
                step=1
            )

            run_button = gr.Button('Generate')
        with gr.Column():
            result = gr.Gallery(label='Result').style(grid=4)
            proxy_words_result = gr.Text(label='Used proxy words')

            examples = [
                [
                    "hamster eating watermelon on the beach",
                    "watermelon",
                    "",
                    20,
                    6,
                    16,
                    "",
                    "hamster,beach",
                    48,
                    None
                ],
                [
                    "A decorated lamp in the livingroom",
                    "lamp",
                    "",
                    20,
                    4,
                    14,
                    "livingroom",
                    "",
                    42,
                    None
                ],
                [
                    "a snake in the field eats an apple",
                    "snake",
                    "",
                    20,
                    7,
                    17,
                    "apple",
                    "apple,field",
                    10,
                    None
                ]
            ]

            gr.Examples(examples=examples,
                        inputs=[
                            prompt,
                            object_of_interest,
                            proxy_words,
                            number_of_variations,
                            start_prompt_range,
                            end_prompt_range,
                            objects_to_preserve,
                            background_nouns,
                            seed,
                            input_image
                        ],
                        outputs=[
                            result,
                            proxy_words_result
                        ],
                        fn=main_pipeline,
                        cache_examples=False)


    inputs = [
        prompt,
        object_of_interest,
        proxy_words,
        number_of_variations,
        start_prompt_range,
        end_prompt_range,
        objects_to_preserve,
        background_nouns,
        seed,
        input_image
    ]
    outputs = [
        result,
        proxy_words_result
    ]
    run_button.click(fn=main_pipeline, inputs=inputs, outputs=outputs)

demo.queue(max_size=50).launch(share=False)