-
Notifications
You must be signed in to change notification settings - Fork 2.3k
/
net_utils.py
223 lines (187 loc) · 7.32 KB
/
net_utils.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
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
import numpy as np
import torchvision.models as models
from model.utils.config import cfg
from model.roi_crop.functions.roi_crop import RoICropFunction
import cv2
import pdb
import random
def save_net(fname, net):
import h5py
h5f = h5py.File(fname, mode='w')
for k, v in net.state_dict().items():
h5f.create_dataset(k, data=v.cpu().numpy())
def load_net(fname, net):
import h5py
h5f = h5py.File(fname, mode='r')
for k, v in net.state_dict().items():
param = torch.from_numpy(np.asarray(h5f[k]))
v.copy_(param)
def weights_normal_init(model, dev=0.01):
if isinstance(model, list):
for m in model:
weights_normal_init(m, dev)
else:
for m in model.modules():
if isinstance(m, nn.Conv2d):
m.weight.data.normal_(0.0, dev)
elif isinstance(m, nn.Linear):
m.weight.data.normal_(0.0, dev)
def clip_gradient(model, clip_norm):
"""Computes a gradient clipping coefficient based on gradient norm."""
totalnorm = 0
for p in model.parameters():
if p.requires_grad:
modulenorm = p.grad.data.norm()
totalnorm += modulenorm ** 2
totalnorm = torch.sqrt(totalnorm).item()
norm = (clip_norm / max(totalnorm, clip_norm))
for p in model.parameters():
if p.requires_grad:
p.grad.mul_(norm)
def vis_detections(im, class_name, dets, thresh=0.8):
"""Visual debugging of detections."""
for i in range(np.minimum(10, dets.shape[0])):
bbox = tuple(int(np.round(x)) for x in dets[i, :4])
score = dets[i, -1]
if score > thresh:
cv2.rectangle(im, bbox[0:2], bbox[2:4], (0, 204, 0), 2)
cv2.putText(im, '%s: %.3f' % (class_name, score), (bbox[0], bbox[1] + 15), cv2.FONT_HERSHEY_PLAIN,
1.0, (0, 0, 255), thickness=1)
return im
def adjust_learning_rate(optimizer, decay=0.1):
"""Sets the learning rate to the initial LR decayed by 0.5 every 20 epochs"""
for param_group in optimizer.param_groups:
param_group['lr'] = decay * param_group['lr']
def save_checkpoint(state, filename):
torch.save(state, filename)
def _smooth_l1_loss(bbox_pred, bbox_targets, bbox_inside_weights, bbox_outside_weights, sigma=1.0, dim=[1]):
sigma_2 = sigma ** 2
box_diff = bbox_pred - bbox_targets
in_box_diff = bbox_inside_weights * box_diff
abs_in_box_diff = torch.abs(in_box_diff)
smoothL1_sign = (abs_in_box_diff < 1. / sigma_2).detach().float()
in_loss_box = torch.pow(in_box_diff, 2) * (sigma_2 / 2.) * smoothL1_sign \
+ (abs_in_box_diff - (0.5 / sigma_2)) * (1. - smoothL1_sign)
out_loss_box = bbox_outside_weights * in_loss_box
loss_box = out_loss_box
for i in sorted(dim, reverse=True):
loss_box = loss_box.sum(i)
loss_box = loss_box.mean()
return loss_box
def _crop_pool_layer(bottom, rois, max_pool=True):
# code modified from
# https://github.com/ruotianluo/pytorch-faster-rcnn
# implement it using stn
# box to affine
# input (x1,y1,x2,y2)
"""
[ x2-x1 x1 + x2 - W + 1 ]
[ ----- 0 --------------- ]
[ W - 1 W - 1 ]
[ ]
[ y2-y1 y1 + y2 - H + 1 ]
[ 0 ----- --------------- ]
[ H - 1 H - 1 ]
"""
rois = rois.detach()
batch_size = bottom.size(0)
D = bottom.size(1)
H = bottom.size(2)
W = bottom.size(3)
roi_per_batch = rois.size(0) / batch_size
x1 = rois[:, 1::4] / 16.0
y1 = rois[:, 2::4] / 16.0
x2 = rois[:, 3::4] / 16.0
y2 = rois[:, 4::4] / 16.0
height = bottom.size(2)
width = bottom.size(3)
# affine theta
zero = Variable(rois.data.new(rois.size(0), 1).zero_())
theta = torch.cat([\
(x2 - x1) / (width - 1),
zero,
(x1 + x2 - width + 1) / (width - 1),
zero,
(y2 - y1) / (height - 1),
(y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)
if max_pool:
pre_pool_size = cfg.POOLING_SIZE * 2
grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, pre_pool_size, pre_pool_size)))
bottom = bottom.view(1, batch_size, D, H, W).contiguous().expand(roi_per_batch, batch_size, D, H, W)\
.contiguous().view(-1, D, H, W)
crops = F.grid_sample(bottom, grid)
crops = F.max_pool2d(crops, 2, 2)
else:
grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, cfg.POOLING_SIZE, cfg.POOLING_SIZE)))
bottom = bottom.view(1, batch_size, D, H, W).contiguous().expand(roi_per_batch, batch_size, D, H, W)\
.contiguous().view(-1, D, H, W)
crops = F.grid_sample(bottom, grid)
return crops, grid
def _affine_grid_gen(rois, input_size, grid_size):
rois = rois.detach()
x1 = rois[:, 1::4] / 16.0
y1 = rois[:, 2::4] / 16.0
x2 = rois[:, 3::4] / 16.0
y2 = rois[:, 4::4] / 16.0
height = input_size[0]
width = input_size[1]
zero = Variable(rois.data.new(rois.size(0), 1).zero_())
theta = torch.cat([\
(x2 - x1) / (width - 1),
zero,
(x1 + x2 - width + 1) / (width - 1),
zero,
(y2 - y1) / (height - 1),
(y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)
grid = F.affine_grid(theta, torch.Size((rois.size(0), 1, grid_size, grid_size)))
return grid
def _affine_theta(rois, input_size):
rois = rois.detach()
x1 = rois[:, 1::4] / 16.0
y1 = rois[:, 2::4] / 16.0
x2 = rois[:, 3::4] / 16.0
y2 = rois[:, 4::4] / 16.0
height = input_size[0]
width = input_size[1]
zero = Variable(rois.data.new(rois.size(0), 1).zero_())
# theta = torch.cat([\
# (x2 - x1) / (width - 1),
# zero,
# (x1 + x2 - width + 1) / (width - 1),
# zero,
# (y2 - y1) / (height - 1),
# (y1 + y2 - height + 1) / (height - 1)], 1).view(-1, 2, 3)
theta = torch.cat([\
(y2 - y1) / (height - 1),
zero,
(y1 + y2 - height + 1) / (height - 1),
zero,
(x2 - x1) / (width - 1),
(x1 + x2 - width + 1) / (width - 1)], 1).view(-1, 2, 3)
return theta
def compare_grid_sample():
# do gradcheck
N = random.randint(1, 8)
C = 2 # random.randint(1, 8)
H = 5 # random.randint(1, 8)
W = 4 # random.randint(1, 8)
input = Variable(torch.randn(N, C, H, W).cuda(), requires_grad=True)
input_p = input.clone().data.contiguous()
grid = Variable(torch.randn(N, H, W, 2).cuda(), requires_grad=True)
grid_clone = grid.clone().contiguous()
out_offcial = F.grid_sample(input, grid)
grad_outputs = Variable(torch.rand(out_offcial.size()).cuda())
grad_outputs_clone = grad_outputs.clone().contiguous()
grad_inputs = torch.autograd.grad(out_offcial, (input, grid), grad_outputs.contiguous())
grad_input_off = grad_inputs[0]
crf = RoICropFunction()
grid_yx = torch.stack([grid_clone.data[:,:,:,1], grid_clone.data[:,:,:,0]], 3).contiguous().cuda()
out_stn = crf.forward(input_p, grid_yx)
grad_inputs = crf.backward(grad_outputs_clone.data)
grad_input_stn = grad_inputs[0]
pdb.set_trace()
delta = (grad_input_off.data - grad_input_stn).sum()