File size: 2,721 Bytes
4d6b877
 
 
 
 
 
 
 
 
 
 
7bf4347
b260763
4d6b877
 
 
 
 
 
 
 
 
 
 
 
b260763
 
 
 
 
 
a813ad5
b260763
 
4d6b877
 
 
 
 
 
 
 
 
 
b260763
 
4d6b877
 
 
 
b260763
4d6b877
b260763
4d6b877
 
 
 
 
 
 
 
b260763
09ef29c
4d6b877
 
 
 
 
 
09ef29c
4d6b877
b260763
 
 
 
 
 
 
7bf4347
 
 
 
 
 
 
4d6b877
 
 
8279921
7bf4347
 
4d6b877
 
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
import os

import torch
import PIL.Image
import numpy as np
import gradio as gr
from yarg import get

from models.stylegan_generator import StyleGANGenerator
from models.stylegan2_generator import StyleGAN2Generator

from utils.constants import VALID_CHOICES, ENABLE_GPU, MODEL_NAME, OUTPUT_LIST, description, title
from utils.image_manip import tensor_to_pil, concat_images

def get_generator(model_name):
    if model_name == 'stylegan_ffhq':
        generator = StyleGANGenerator(model_name)
    elif model_name == 'stylegan2_ffhq':
        generator = StyleGAN2Generator(model_name)
    else:
        raise ValueError('Model name not recognized')
    if ENABLE_GPU:
        generator = generator.cuda()
    return generator

generator = get_generator(MODEL_NAME)
boundaries = {
    boundary:np.squeeze(np.load(open(os.path.join('boundaries', MODEL_NAME, 'boundary_%s.npy' % boundary), 'rb'))) 
    for boundary in VALID_CHOICES
}

@torch.no_grad()
def inference(seed, coef, nb_images, list_choices):
    global generator, boundaries
    np.random.seed(seed)
    latent_codes = generator.easy_sample(nb_images)
    if ENABLE_GPU:
        latent_codes = latent_codes.cuda()
        generator = generator.cuda()
    generated_images = generator.easy_synthesize(latent_codes)
    generated_images = tensor_to_pil(generated_images)

    new_latent_codes = latent_codes.copy()
    for i, _ in enumerate(generated_images):
        for choice in list_choices:
            new_latent_codes[i, :] +=  boundaries[choice]*coef

    modified_generated_images = generator.easy_synthesize(new_latent_codes)
    modified_generated_images = tensor_to_pil(modified_generated_images)

    concatenated_output = concat_images(generated_images, modified_generated_images)

    return concatenated_output

iface = gr.Interface(
    fn=inference, 
    inputs=[
        gr.inputs.Slider(
            minimum=0,
            maximum=1000,
            step=1,
            default=644,
            label="Random seed to use for the generation"
        ),
        gr.inputs.Slider(
            minimum=-3,
            maximum=3,
            step=0.1,
            default=0,
            label="Modification coefficient",
        ),
        gr.inputs.Slider(
            minimum=1,
            maximum=10,
            step=1,
            default=2,
            label="Number of images to generate",
        ),
        gr.inputs.CheckboxGroup(
            VALID_CHOICES, 
            default=[], 
            type="value", 
            label="Select attributes to modify", 
            optional=False
        )
    ],
    outputs=OUTPUT_LIST,
    layout="horizontal",
    theme="peach",
    description=description,
    title=title,
)
iface.launch()