FaceSwapPipeline/upsampler.py at main · postworthy/FaceSwapPipeline · GitHub

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
import os
import cv2
os.environ['OMP_NUM_THREADS'] = '1'
import threading
from realesrgan import RealESRGANer
from realesrgan.archs.srvgg_arch import SRVGGNetCompact
from basicsr.archs.rrdbnet_arch import RRDBNet
from gfpgan import GFPGANer #https://github.com/postworthy/GFPGAN


THREAD_LOCK_UPSAMPLER = threading.Lock()
THREAD_LOCK_UPSAMPLER_FAST = threading.Lock()
THREAD_LOCK_PROCESS = threading.Lock()
UPSAMPLER_BG = None
UPSAMPLER = {"2":None, "4":None}
UPSAMPLER_FAST = None

def get_full_upsampler(up_by=4):
    #https://github.com/xinntao/Real-ESRGAN/blob/master/inference_realesrgan.py
    global UPSAMPLER
    if UPSAMPLER[str(up_by)] == None:
        with THREAD_LOCK_UPSAMPLER:
            if UPSAMPLER[str(up_by)] == None:
                bg_model_path = os.path.join(os.path.abspath(os.path.dirname(__file__)), 'RealESRGAN_x4plus.pth')
                face_model_path = os.path.join(os.path.abspath(os.path.dirname(__file__)), 'GFPGANv1.4.pth')
                upsampler = RealESRGANer(
                    scale=up_by,
                    model_path=bg_model_path,
                    dni_weight=None,
                    model=RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4),
                    tile=0,
                    tile_pad=10,
                    pre_pad=0,
                    half=False,
                    gpu_id=0
                )
                face_upsampler = GFPGANer(
                    model_path=face_model_path,
                    upscale=up_by,
                    arch='clean',
                    channel_multiplier=2,
                    bg_upsampler=upsampler
                )
                UPSAMPLER[str(up_by)] = face_upsampler
    return UPSAMPLER[str(up_by)]

def get_bg_upsampler(upscale=4):
    #https://github.com/xinntao/Real-ESRGAN/blob/master/inference_realesrgan.py
    global UPSAMPLER_BG
    if not UPSAMPLER_BG:
        with THREAD_LOCK_UPSAMPLER:
            if not UPSAMPLER_BG:
                bg_model_path = os.path.join(os.path.abspath(os.path.dirname(__file__)), 'RealESRGAN_x4plus.pth')
                upsampler = RealESRGANer(
                    scale=upscale,
                    model_path=bg_model_path,
                    dni_weight=None,
                    model=RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64, num_block=23, num_grow_ch=32, scale=4),
                    tile=0,
                    tile_pad=10,
                    pre_pad=0,
                    half=False,
                    gpu_id=0
                )
                UPSAMPLER_BG = upsampler
    return UPSAMPLER_BG

def get_face_upsampler(upscale=2, bg_upsampler=None):
    #https://github.com/xinntao/Real-ESRGAN/blob/master/inference_realesrgan.py
    global UPSAMPLER_FAST
    if not UPSAMPLER_FAST:
        with THREAD_LOCK_UPSAMPLER_FAST:
            if not UPSAMPLER_FAST:
                face_model_path = os.path.join(os.path.abspath(os.path.dirname(__file__)), 'GFPGANv1.4.pth')
                face_upsampler = GFPGANer(
                    model_path=face_model_path,
                    upscale=upscale,
                    arch='clean',
                    channel_multiplier=2,
                    bg_upsampler=bg_upsampler
                )
                UPSAMPLER_FAST = face_upsampler
    return UPSAMPLER_FAST

def get_upsampler(fast=True, up_by=4):
    return get_face_upsampler(up_by) if fast else get_full_upsampler(up_by)

def upsample(image_data, fast=True, has_aligned=False, up_by=None):
    from util import get_face_analyser
    from masks import get_final_image, merge_original
    from insightface.utils import face_align

    if up_by == None and fast:
        up_by = 2
    elif up_by == None:
        up_by = 4

    #if fast:
    upsampler = get_upsampler(fast, up_by=up_by)

    with THREAD_LOCK_PROCESS:
        _, _, output = upsampler.enhance(image_data, has_aligned=has_aligned, only_center_face=False, paste_back=True)

        upsampler.cleanup() #Requires using https://github.com/postworthy/GFPGAN

    return output
    #else:
        #bg_upsampler = get_bg_upsampler(4)
        #face_upsampler = get_face_upsampler(4)
        #face_analyser = get_face_analyser()

        #with THREAD_LOCK_PROCESS:
        #    bg_img = bg_upsampler.enhance(image_data)[0]
        #    face = face_analyser.get(bg_img)[0]
        #    warped_img, estimated_norm = face_align.norm_crop2(bg_img, face.kps, 512)
        #    _, restored_faces, _ = face_upsampler.enhance(image_data, has_aligned=has_aligned, only_center_face=False, paste_back=False)
        #    #output = get_final_image([[restored_faces[0]]], bg_img, [[estimated_norm]])
        #    output = merge_original(bg_img, restored_faces[0], warped_img, estimated_norm)

        #    #Requires using https://github.com/postworthy/GFPGAN
        #    ##bg_upsampler.cleanup()
        #    face_upsampler.cleanup()

        #return output

def upsample_batch(images:list, fast=True, has_aligned=False, up_by=None):
    from util import get_face_analyser
    from masks import get_final_image, merge_original

    if up_by is None and fast:
        up_by = 2
    elif up_by is None:
        up_by = 4

    upsampler = get_upsampler(fast, up_by=up_by)

    processed_images = []
    with THREAD_LOCK_PROCESS:
        for image_data in images:
            _, _, output = upsampler.enhance(image_data, has_aligned=has_aligned, only_center_face=False, paste_back=True)
            processed_images.append(output)

        upsampler.cleanup()  # Requires using https://github.com/postworthy/GFPGAN

    return processed_images


    return results