/
model.py
384 lines (279 loc) · 12 KB
/
model.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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
import torch
import torch.nn as nn
import torch.nn.functional as F
from networks.resnet_encoder import ResnetEncoder
from networks.depth_decoder import DepthDecoder
from networks.pose_decoder import PoseDecoder
from layers import BackprojectDepth, Project3D, get_smooth_loss, transformation_from_parameters, disp_to_depth
from loss import SSIM
class MotionNet(nn.Module):
def __init__(self):
super(MotionNet, self).__init__()
self.visual_encoder = ResnetEncoder(34, True, num_input_images=1)
self.motion_decoder = PoseDecoder(self.visual_encoder.num_ch_enc, num_input_features=1, num_frames_to_predict_for=2)
def forward(self, inputs):
motion_inputs1 = inputs["color_aug", -1, 0]
motion_inputs2 = inputs["color_aug", 0, 0]
enc1 = self.visual_encoder(motion_inputs1, normalize=True)
enc2 = self.visual_encoder(motion_inputs2, normalize=True)
axisangle1, translation1 = self.motion_decoder([enc1])
axisangle2, translation2 = self.motion_decoder([enc2])
return axisangle1, translation1, axisangle2, translation2
class Monodepth(nn.Module):
def __init__(self, stage = 1, batch_size=1):
super(Monodepth, self).__init__()
self.stage = stage
self.num_scales = len([0, 1, 2, 3])
self.scales = [0, 1, 2, 3]
self.frame_ids = [0, -1, 1]
self.height = 160
self.width = 320
self.num_input_frames = len([0, -1, 1])
self.num_pose_frames = 2
self.min_depth = 0.1
self.max_depth = 100.0
self.depth_encoder = ResnetEncoder(18, True)
self.depth_decoder = DepthDecoder(self.depth_encoder.num_ch_enc, self.scales)
self.pose_encoder = ResnetEncoder(18, True, num_input_images=self.num_pose_frames)
self.pose_decoder = PoseDecoder(self.pose_encoder.num_ch_enc, num_input_features=1, num_frames_to_predict_for=2)
self.fl = nn.Sequential(nn.Linear(512, 256),
nn.ReLU(True),
nn.Linear(256, 2),
nn.Softplus()
)
self.offset = nn.Sequential(nn.Linear(512, 256),
nn.ReLU(True),
nn.Linear(256, 2),
nn.Sigmoid()
)
self.avg_pooling = nn.AdaptiveAvgPool2d((1,1))
self.backproject_depth = {}
self.project_3d = {}
for scale in self.scales:
h = self.height // (2 ** scale)
w = self.width // (2 ** scale)
self.backproject_depth[scale] = BackprojectDepth(batch_size, h, w)
self.project_3d[scale] = Project3D(batch_size, h, w)
self.ssim = SSIM()
self.initialized = False
def initialize(self):
for scale in self.scales:
self.backproject_depth[scale] = self.backproject_depth[scale].to(self.device)
self.project_3d[scale] = self.project_3d[scale].to(self.device)
self.initialized = True
def forward_stage1(self, inputs):
if not self.initialized:
self.device = inputs["color_aug", 0, 0].device
self.initialize()
# depth prediction
features = self.depth_encoder(inputs["color_aug", 0, 0], normalize=True)
outputs = self.depth_decoder(features)
pose_feats = {f_i: inputs["color_aug", f_i, 0] for f_i in self.frame_ids}
poses_inputs1 = torch.cat([pose_feats[-1], pose_feats[0]], 1)
poses_inputs2 = torch.cat([pose_feats[0], pose_feats[1]], 1)
pose_enc1 = self.pose_encoder(poses_inputs1, normalize=True)
pose_enc2 = self.pose_encoder(poses_inputs2, normalize=True)
# intrinsics prediction
feature_pooled1 = torch.flatten(self.avg_pooling(pose_enc1[-1]), 1)
fl1 = self.fl(feature_pooled1)
offsets1 = self.offset(feature_pooled1)
feature_pooled2 = torch.flatten(self.avg_pooling(pose_enc2[-1]), 1)
fl2 = self.fl(feature_pooled2)
offsets2 = self.offset(feature_pooled2)
K1 = self.compute_K(fl1, offsets1)
K2 = self.compute_K(fl2, offsets2)
K = (K1+K2)/2
inputs = self.add_K(K, inputs)
# pose prediction
axisangle1, translation1 = self.pose_decoder([pose_enc1])
axisangle2, translation2 = self.pose_decoder([pose_enc2])
outputs[("axisangle", 0, -1)] = axisangle1
outputs[("translation", 0, -1)] = translation1
outputs[("axisangle", 0, 1)] = axisangle2
outputs[("translation", 0, 1)] = translation2
outputs[("cam_T_cam", 0, -1)] = transformation_from_parameters(
axisangle1[:, 0], translation1[:, 0], invert=True)
outputs[("cam_T_cam", 0, 1)] = transformation_from_parameters(
axisangle2[:, 0], translation2[:, 0], invert=False)
for scale in self.scales:
disp = outputs[("disp", scale)]
disp = F.interpolate(
disp, [self.height, self.width], mode="bilinear", align_corners=False)
source_scale = 0
_, depth = disp_to_depth(disp, self.min_depth, self.max_depth)
outputs[("depth", 0, scale)] = depth
for i, frame_id in enumerate(self.frame_ids[1:]):
T = outputs[("cam_T_cam", 0, frame_id)]
cam_points = self.backproject_depth[source_scale](
depth, inputs[("inv_K", source_scale)])
pix_coords = self.project_3d[source_scale](
cam_points, inputs[("K", source_scale)], T)
outputs[("sample", frame_id, scale)] = pix_coords
outputs[("color", frame_id, scale)] = F.grid_sample(
inputs[("color", frame_id, source_scale)],
outputs[("sample", frame_id, scale)],
padding_mode="border")
outputs[("color_identity", frame_id, scale)] = \
inputs[("color", frame_id, source_scale)]
losses = {}
total_loss = 0
for scale in self.scales:
loss = 0
reprojection_losses = []
source_scale = 0
disp = outputs[("disp", scale)]
color = inputs[("color", 0, scale)]
target = inputs[("color", 0, source_scale)]
for frame_id in self.frame_ids[1:]:
pred = outputs[("color", frame_id, scale)]
reprojection_losses.append(self.compute_reprojection_loss(pred, target))
reprojection_losses = torch.cat(reprojection_losses, 1)
identity_reprojection_losses = []
for frame_id in self.frame_ids[1:]:
pred = inputs[("color", frame_id, source_scale)]
identity_reprojection_losses.append(
self.compute_reprojection_loss(pred, target))
identity_reprojection_losses = torch.cat(identity_reprojection_losses, 1)
identity_reprojection_loss = identity_reprojection_losses
reprojection_loss = reprojection_losses
identity_reprojection_loss += torch.randn(
identity_reprojection_loss.shape, device=self.device) * 0.00001
combined = torch.cat((identity_reprojection_loss, reprojection_loss), dim=1)
if combined.shape[1] == 1:
to_optimise = combined
else:
to_optimise, idxs = torch.min(combined, dim=1)
outputs["identity_selection/{}".format(scale)] = (
idxs > identity_reprojection_loss.shape[1] - 1).float()
loss += to_optimise.mean()
mean_disp = disp.mean(2, True).mean(3, True)
norm_disp = disp / (mean_disp + 1e-7)
smooth_loss = get_smooth_loss(norm_disp, color)
loss += 1e-3 * smooth_loss / (2 ** scale)
total_loss += loss
losses["loss/{}".format(scale)] = loss
total_loss /= self.num_scales
losses["loss"] = total_loss
return outputs, losses
def forward_stage2(self, inputs, axisangle1, translation1, axisangle2, translation2):
if not self.initialized:
self.device = inputs["color_aug", 0, 0].device
self.initialize()
with torch.no_grad():
features = self.depth_encoder(inputs["color", 0, 0], normalize=True)
outputs = self.depth_decoder(features)
pose_feats = {f_i: inputs["color", f_i, 0] for f_i in self.frame_ids}
poses_inputs1 = torch.cat([pose_feats[-1], pose_feats[0]], 1)
poses_inputs2 = torch.cat([pose_feats[0], pose_feats[1]], 1)
pose_enc1 = self.pose_encoder(poses_inputs1, normalize=True)
pose_enc2 = self.pose_encoder(poses_inputs2, normalize=True)
feature_pooled1 = torch.flatten(self.avg_pooling(pose_enc1[-1]), 1)
fl1 = self.fl(feature_pooled1)
offsets1 = self.offset(feature_pooled1)
feature_pooled2 = torch.flatten(self.avg_pooling(pose_enc2[-1]), 1)
fl2 = self.fl(feature_pooled2)
offsets2 = self.offset(feature_pooled2)
K1 = self.compute_K(fl1, offsets1)
K2 = self.compute_K(fl2, offsets2)
K = (K1+K2)/2
inputs = self.add_K(K, inputs)
outputs[("axisangle", 0, -1)] = axisangle1
outputs[("translation", 0, -1)] = translation1
outputs[("axisangle", 0, 1)] = axisangle2
outputs[("translation", 0, 1)] = translation2
outputs[("cam_T_cam", 0, -1)] = transformation_from_parameters(
axisangle1[:, 0], translation1[:, 0], invert=True)
outputs[("cam_T_cam", 0, 1)] = transformation_from_parameters(
axisangle2[:, 0], translation2[:, 0], invert=False)
for scale in [0]:
with torch.no_grad():
disp = outputs[("disp", scale)]
disp = F.interpolate(
disp, [self.height, self.width], mode="bilinear", align_corners=False)
source_scale = 0
_, depth = disp_to_depth(disp, self.min_depth, self.max_depth)
outputs[("depth", 0, scale)] = depth
for i, frame_id in enumerate(self.frame_ids[1:]):
T = outputs[("cam_T_cam", 0, frame_id)]
cam_points = self.backproject_depth[source_scale](
depth, inputs[("inv_K", source_scale)])
pix_coords = self.project_3d[source_scale](
cam_points, inputs[("K", source_scale)], T)
outputs[("sample", frame_id, scale)] = pix_coords
outputs[("color", frame_id, scale)] = F.grid_sample(
inputs[("color", frame_id, source_scale)],
outputs[("sample", frame_id, scale)],
padding_mode="border")
outputs[("color_identity", frame_id, scale)] = \
inputs[("color", frame_id, source_scale)]
losses = {}
total_loss = 0
for scale in [0]:
loss = 0
reprojection_losses = []
source_scale = 0
disp = outputs[("disp", scale)]
color = inputs[("color", 0, scale)]
target = inputs[("color", 0, source_scale)]
for frame_id in self.frame_ids[1:]:
pred = outputs[("color", frame_id, scale)]
reprojection_losses.append(self.compute_reprojection_loss(pred, target))
reprojection_losses = torch.cat(reprojection_losses, 1)
identity_reprojection_losses = []
for frame_id in self.frame_ids[1:]:
pred = inputs[("color", frame_id, source_scale)]
identity_reprojection_losses.append(
self.compute_reprojection_loss(pred, target))
identity_reprojection_losses = torch.cat(identity_reprojection_losses, 1)
identity_reprojection_loss = identity_reprojection_losses
reprojection_loss = reprojection_losses
identity_reprojection_loss += torch.randn(
identity_reprojection_loss.shape, device=self.device) * 0.00001
combined = torch.cat((identity_reprojection_loss, reprojection_loss), dim=1)
if combined.shape[1] == 1:
to_optimise = combined
else:
to_optimise, idxs = torch.min(combined, dim=1)
outputs["identity_selection/{}".format(scale)] = (
idxs > identity_reprojection_loss.shape[1] - 1).float()
loss += to_optimise.mean()
mean_disp = disp.mean(2, True).mean(3, True)
norm_disp = disp / (mean_disp + 1e-7)
smooth_loss = get_smooth_loss(norm_disp, color)
loss += 1e-3 * smooth_loss / (2 ** scale)
total_loss += loss
losses["loss/{}".format(scale)] = loss
total_loss /= self.num_scales
losses["loss"] = total_loss
return outputs, losses
def forward(self, inputs, axisangle1=None, translation1=None, axisangle2=None, translation2=None):
if self.stage == 1:
return self.forward_stage1(inputs)
else:
return self.forward_stage2(inputs, axisangle1, translation1, axisangle2, translation2)
def compute_reprojection_loss(self, pred, target):
"""Computes reprojection loss between a batch of predicted and target images
"""
abs_diff = torch.abs(target - pred)
l1_loss = abs_diff.mean(1, True)
ssim_loss = self.ssim(pred, target).mean(1, True)
reprojection_loss = 0.85 * ssim_loss + 0.15 * l1_loss
return reprojection_loss
def compute_K(self, fl, offsets):
B = fl.shape[0]
fl = torch.diag_embed(fl) # B * 2 * 2
K = torch.cat([fl, offsets.view(-1, 2, 1)], 2) # B * 2 * 3
row = torch.tensor([[0, 0, 1], [0, 0, 0]]).view(1, 2, 3).repeat(B, 1, 1).type_as(K)
K = torch.cat([K, row], 1) # B * 4 * 3
col = torch.tensor([0, 0, 0, 1]).view(1, 4, 1).repeat(B, 1, 1).type_as(K)
K = torch.cat([K, col], 2) # B * 4 * 4
return K
def add_K(self, K, inputs):
for scale in self.scales:
K_scale = K.clone()
K_scale[:, 0] *= self.width // (2 ** scale)
K_scale[:, 1] *= self.height // (2 ** scale)
inv_K_scale = torch.linalg.pinv(K_scale)
inputs[("K", scale)] = K_scale
inputs[("inv_K", scale)] = inv_K_scale
return inputs