-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbuild_sam3D.py
218 lines (200 loc) · 6.48 KB
/
build_sam3D.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
218
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
import torch
from functools import partial
from .modeling import ImageEncoderViT3D, MaskDecoder3D, PromptEncoder3D, Sam3D
def build_sam3D_vit_h(checkpoint=None):
return _build_sam3D(
encoder_embed_dim=1280,
encoder_depth=32,
encoder_num_heads=16,
encoder_global_attn_indexes=[7, 15, 23, 31],
checkpoint=checkpoint,
)
build_sam3D = build_sam3D_vit_h
def build_sam3D_vit_l(checkpoint=None):
return _build_sam3D(
encoder_embed_dim=1024,
encoder_depth=24,
encoder_num_heads=16,
encoder_global_attn_indexes=[5, 11, 17, 23],
checkpoint=checkpoint,
)
def build_sam3D_vit_b(checkpoint=None):
return _build_sam3D(
# encoder_embed_dim=768,
encoder_embed_dim=384,
encoder_depth=12,
encoder_num_heads=12,
encoder_global_attn_indexes=[2, 5, 8, 11],
checkpoint=checkpoint,
)
def build_sam3D_vit_b_ori(checkpoint=None):
return _build_sam3D_fast(
encoder_embed_dim=768,
encoder_depth=12,
encoder_num_heads=12,
encoder_global_attn_indexes=[2, 5, 8, 11],
checkpoint=checkpoint,
)
def build_sam3D_vit_b_original(checkpoint=None):
return _build_sam3D_ori(
encoder_embed_dim=768,
encoder_depth=12,
encoder_num_heads=12,
encoder_global_attn_indexes=[2, 5, 8, 11],
checkpoint=checkpoint,
)
sam_model_registry3D = {
"default": build_sam3D_vit_h,
"vit_h": build_sam3D_vit_h,
"vit_l": build_sam3D_vit_l,
"vit_b": build_sam3D_vit_b,
"vit_b_ori": build_sam3D_vit_b_ori,
"vit_b_original":build_sam3D_vit_b_original,
}
def _build_sam3D(
encoder_embed_dim,
encoder_depth,
encoder_num_heads,
encoder_global_attn_indexes,
checkpoint=None,
):
prompt_embed_dim = 384
image_size = 256 #
vit_patch_size = 16
image_embedding_size = image_size // vit_patch_size
sam = Sam3D(
image_encoder=ImageEncoderViT3D(
depth=encoder_depth,
embed_dim=encoder_embed_dim,
img_size=image_size,
mlp_ratio=4,
norm_layer=partial(torch.nn.LayerNorm, eps=1e-6),
num_heads=encoder_num_heads,
patch_size=vit_patch_size,
qkv_bias=True,
use_rel_pos=True,
global_attn_indexes=encoder_global_attn_indexes,
window_size=14,
out_chans=prompt_embed_dim,
),
prompt_encoder=PromptEncoder3D(
embed_dim=prompt_embed_dim,
image_embedding_size=(image_embedding_size, image_embedding_size, image_embedding_size),
input_image_size=(image_size, image_size, image_size),
mask_in_chans=16,
),
mask_decoder=MaskDecoder3D(
num_multimask_outputs=3,
transformer_dim=prompt_embed_dim,
iou_head_depth=3,
iou_head_hidden_dim=256,
),
pixel_mean=[123.675, 116.28, 103.53],
pixel_std=[58.395, 57.12, 57.375],
)
sam.eval()
if checkpoint is not None:
with open(checkpoint, "rb") as f:
state_dict = torch.load(f)
sam.load_state_dict(state_dict)
return sam
def _build_sam3D_ori(
encoder_embed_dim,
encoder_depth,
encoder_num_heads,
encoder_global_attn_indexes,
checkpoint=None,
):
prompt_embed_dim = 384
image_size = 128
vit_patch_size = 16
image_embedding_size = image_size // vit_patch_size
sam = Sam3D(
image_encoder=ImageEncoderViT3D(
depth=encoder_depth,
embed_dim=encoder_embed_dim,
img_size=image_size,
mlp_ratio=4,
norm_layer=partial(torch.nn.LayerNorm, eps=1e-6),
num_heads=encoder_num_heads,
patch_size=vit_patch_size,
qkv_bias=True,
use_rel_pos=True,
global_attn_indexes=encoder_global_attn_indexes,
window_size=14,
out_chans=prompt_embed_dim,
skip_layer = 0,
),
prompt_encoder=PromptEncoder3D(
embed_dim=prompt_embed_dim,
image_embedding_size=(image_embedding_size, image_embedding_size, image_embedding_size),
input_image_size=(image_size, image_size, image_size),
mask_in_chans=16,
),
mask_decoder=MaskDecoder3D(
num_multimask_outputs=3,
transformer_dim=prompt_embed_dim,
iou_head_depth=3,
iou_head_hidden_dim=256,
),
pixel_mean=[123.675, 116.28, 103.53],
pixel_std=[58.395, 57.12, 57.375],
)
sam.eval()
if checkpoint is not None:
with open(checkpoint, "rb") as f:
state_dict = torch.load(f)
sam.load_state_dict(state_dict)
return sam
def _build_sam3D_fast(
encoder_embed_dim,
encoder_depth,
encoder_num_heads,
encoder_global_attn_indexes,
checkpoint=None,
):
prompt_embed_dim = 384
image_size = 128
vit_patch_size = 16
image_embedding_size = image_size // vit_patch_size
sam = Sam3D(
image_encoder=ImageEncoderViT3D(
depth=encoder_depth,
embed_dim=encoder_embed_dim,
img_size=image_size,
mlp_ratio=4,
norm_layer=partial(torch.nn.LayerNorm, eps=1e-6),
num_heads=encoder_num_heads,
patch_size=vit_patch_size,
qkv_bias=True,
use_rel_pos=True,
global_attn_indexes=encoder_global_attn_indexes,
window_size=14,
out_chans=prompt_embed_dim,
skip_layer = 0, # default is 0.
),
prompt_encoder=PromptEncoder3D(
embed_dim=prompt_embed_dim,
image_embedding_size=(image_embedding_size, image_embedding_size, image_embedding_size),
input_image_size=(image_size, image_size, image_size),
mask_in_chans=16,
),
mask_decoder=MaskDecoder3D(
num_multimask_outputs=3,
transformer_dim=prompt_embed_dim,
iou_head_depth=3,
iou_head_hidden_dim=256,
),
pixel_mean=[123.675, 116.28, 103.53],
pixel_std=[58.395, 57.12, 57.375],
)
sam.eval()
if checkpoint is not None:
with open(checkpoint, "rb") as f:
state_dict = torch.load(f)
sam.load_state_dict(state_dict)
return sam