File size: 6,979 Bytes
515f781
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# based on https://github.com/isl-org/MiDaS

import cv2
import torch
import torch.nn as nn
import os
models_path = 'pretrained/controlnet/preprocess'

from torchvision.transforms import Compose

from .midas.dpt_depth import DPTDepthModel
from .midas.midas_net import MidasNet
from .midas.midas_net_custom import MidasNet_small
from .midas.transforms import Resize, NormalizeImage, PrepareForNet

base_model_path = os.path.join(models_path, "midas")
old_modeldir = os.path.dirname(os.path.realpath(__file__))
remote_model_path = "https://huggingface.co/lllyasviel/ControlNet/resolve/main/annotator/ckpts/dpt_hybrid-midas-501f0c75.pt"

ISL_PATHS = {
    "dpt_large": os.path.join(base_model_path, "dpt_large-midas-2f21e586.pt"),
    "dpt_hybrid": os.path.join(base_model_path, "dpt_hybrid-midas-501f0c75.pt"),
    "midas_v21": "",
    "midas_v21_small": "",
}

OLD_ISL_PATHS = {
    "dpt_large": os.path.join(old_modeldir, "dpt_large-midas-2f21e586.pt"),
    "dpt_hybrid": os.path.join(old_modeldir, "dpt_hybrid-midas-501f0c75.pt"),
    "midas_v21": "",
    "midas_v21_small": "",
}


def disabled_train(self, mode=True):
    """Overwrite model.train with this function to make sure train/eval mode
    does not change anymore."""
    return self


def load_midas_transform(model_type):
    # https://github.com/isl-org/MiDaS/blob/master/run.py
    # load transform only
    if model_type == "dpt_large":  # DPT-Large
        net_w, net_h = 384, 384
        resize_mode = "minimal"
        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])

    elif model_type == "dpt_hybrid":  # DPT-Hybrid
        net_w, net_h = 384, 384
        resize_mode = "minimal"
        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])

    elif model_type == "midas_v21":
        net_w, net_h = 384, 384
        resize_mode = "upper_bound"
        normalization = NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

    elif model_type == "midas_v21_small":
        net_w, net_h = 256, 256
        resize_mode = "upper_bound"
        normalization = NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

    else:
        assert False, f"model_type '{model_type}' not implemented, use: --model_type large"

    transform = Compose(
        [
            Resize(
                net_w,
                net_h,
                resize_target=None,
                keep_aspect_ratio=True,
                ensure_multiple_of=32,
                resize_method=resize_mode,
                image_interpolation_method=cv2.INTER_CUBIC,
            ),
            normalization,
            PrepareForNet(),
        ]
    )

    return transform


def load_file_from_url(url, model_dir=None, progress=True, file_name=None):
    """Load file form http url, will download models if necessary.

    Ref:https://github.com/1adrianb/face-alignment/blob/master/face_alignment/utils.py

    Args:
        url (str): URL to be downloaded.
        model_dir (str): The path to save the downloaded model. Should be a full path. If None, use pytorch hub_dir.
            Default: None.
        progress (bool): Whether to show the download progress. Default: True.
        file_name (str): The downloaded file name. If None, use the file name in the url. Default: None.

    Returns:
        str: The path to the downloaded file.
    """
    from torch.hub import download_url_to_file, get_dir
    from urllib.parse import urlparse
    if model_dir is None:  # use the pytorch hub_dir
        hub_dir = get_dir()
        model_dir = os.path.join(hub_dir, 'checkpoints')

    os.makedirs(model_dir, exist_ok=True)

    parts = urlparse(url)
    filename = os.path.basename(parts.path)
    if file_name is not None:
        filename = file_name
    cached_file = os.path.abspath(os.path.join(model_dir, filename))
    if not os.path.exists(cached_file):
        print(f'Downloading: "{url}" to {cached_file}\n')
        download_url_to_file(url, cached_file, hash_prefix=None, progress=progress)
    return cached_file


def load_model(model_type):
    # https://github.com/isl-org/MiDaS/blob/master/run.py
    # load network
    model_path = ISL_PATHS[model_type]
    old_model_path = OLD_ISL_PATHS[model_type]
    if model_type == "dpt_large":  # DPT-Large
        model = DPTDepthModel(
            path=model_path,
            backbone="vitl16_384",
            non_negative=True,
        )
        net_w, net_h = 384, 384
        resize_mode = "minimal"
        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])

    elif model_type == "dpt_hybrid":  # DPT-Hybrid
        if os.path.exists(old_model_path):
            model_path = old_model_path
        elif not os.path.exists(model_path):
            load_file_from_url(remote_model_path, model_dir=base_model_path)
        
        model = DPTDepthModel(
            path=model_path,
            backbone="vitb_rn50_384",
            non_negative=True,
        )
        net_w, net_h = 384, 384
        resize_mode = "minimal"
        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])

    elif model_type == "midas_v21":
        model = MidasNet(model_path, non_negative=True)
        net_w, net_h = 384, 384
        resize_mode = "upper_bound"
        normalization = NormalizeImage(
            mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
        )

    elif model_type == "midas_v21_small":
        model = MidasNet_small(model_path, features=64, backbone="efficientnet_lite3", exportable=True,
                               non_negative=True, blocks={'expand': True})
        net_w, net_h = 256, 256
        resize_mode = "upper_bound"
        normalization = NormalizeImage(
            mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
        )

    else:
        print(f"model_type '{model_type}' not implemented, use: --model_type large")
        assert False

    transform = Compose(
        [
            Resize(
                net_w,
                net_h,
                resize_target=None,
                keep_aspect_ratio=True,
                ensure_multiple_of=32,
                resize_method=resize_mode,
                image_interpolation_method=cv2.INTER_CUBIC,
            ),
            normalization,
            PrepareForNet(),
        ]
    )

    return model.eval(), transform


class MiDaSInference(nn.Module):
    MODEL_TYPES_TORCH_HUB = [
        "DPT_Large",
        "DPT_Hybrid",
        "MiDaS_small"
    ]
    MODEL_TYPES_ISL = [
        "dpt_large",
        "dpt_hybrid",
        "midas_v21",
        "midas_v21_small",
    ]

    def __init__(self, model_type):
        super().__init__()
        assert (model_type in self.MODEL_TYPES_ISL)
        model, _ = load_model(model_type)
        self.model = model
        self.model.train = disabled_train

    def forward(self, x):
        with torch.no_grad():
            prediction = self.model(x)
        return prediction