forked from lxg2015/faceboxes
-
Notifications
You must be signed in to change notification settings - Fork 3
/
multibox_loss.py
101 lines (88 loc) · 3.4 KB
/
multibox_loss.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
#encoding:utf-8
import math
import torch
import torch.nn as nn
import torch.nn.init as init
import torch.nn.functional as F
from torch.autograd import Variable
class MultiBoxLoss(nn.Module):
num_classes = 2
def __init__(self):
super(MultiBoxLoss,self).__init__()
def cross_entropy_loss(self, x, y):
x = x.detach()
y = y.detach()
xmax = x.data.max()
#xmax = xmax.detach()
log_sum_exp = torch.log(torch.sum(torch.exp(x-xmax), 1, keepdim=True)) + xmax
return log_sum_exp - x.gather(1, y.view(-1,1))
def hard_negative_mining(self,conf_loss,pos):
'''
conf_loss [N*21482,]
pos [N,21482]
return negative indice
'''
batch_size,num_boxes = pos.size()
conf_loss[pos.view(-1,1)] = 0 #去掉正样本,the rest are neg conf_loss
conf_loss = conf_loss.view(batch_size,-1)
_,idx = conf_loss.sort(1,descending=True)
_,rank = idx.sort(1)
num_pos = pos.long().sum(1,keepdim=True)
num_neg = torch.clamp(3*num_pos, max=num_boxes-1)
neg = rank < num_neg.expand_as(rank)
return neg
def forward(self,loc_preds,loc_targets,conf_preds,conf_targets):
'''
loc_preds[batch,21842,4]
loc_targets[batch,21842,4]
conf_preds[batch,21842,2]
conf_targets[batch,21842]
'''
batch_size,num_boxes, _ = loc_preds.size()
#print(batch_size,num_boxes)
#print('ok1')
pos = conf_targets>0 #大于0的地方，说明匹配到了人脸框
num_pos = pos.long().sum(1, keepdim=True)
#print(torch.sum(pos))
#print(conf_targets.size())
num_matched_boxes = pos.data.long().sum()
if num_matched_boxes == 0:
return Variable(torch.Tensor([0]),requires_grad=True)
#print('ok2')
pos_mask1 = pos.unsqueeze(2).expand_as(loc_preds)
#print(pos_mask1.size())
# print('pos_mask1 sum {}'.format(torch.sum(pos_mask1)))
pos_loc_preds = loc_preds[pos_mask1].view(-1,4)
pos_loc_targets = loc_targets[pos_mask1].view(-1,4)
# for i in range(num_matched_boxes):
# print(i, pos_loc_preds[i,:], pos_loc_targets[i,:])
loc_loss = F.smooth_l1_loss(pos_loc_preds,pos_loc_targets,size_average=False)
# if loc_loss.data[0] > 10000: #是因为preds有非常大的，导致loss很大，所以是正常的。
# print('preds', pos_loc_preds)
# print('targets', pos_loc_targets)
#print('ok3')
#temp_conf_loss = Variable(requires_grad=False)
conf_loss = self.cross_entropy_loss(conf_preds.view(-1,self.num_classes),
conf_targets.view(-1,1))
#print('conf_loss size {}'.format(conf_loss.size()))
neg = self.hard_negative_mining(conf_loss, pos)
pos_mask = pos.unsqueeze(2).expand_as(conf_preds)
neg_mask = neg.unsqueeze(2).expand_as(conf_preds)
# print('sum neg mask {} size {}'.format(torch.sum(neg_mask),neg_mask.size()))
# print('sum pos mask {} size {}'.format(torch.sum(pos_mask),pos_mask.size()))
#print(neg_mask)
mask = (pos_mask+neg_mask).gt(0)
# print('sum mask {} size {}'.format(torch.sum(mask),mask.size()))
pos_and_neg = (pos+neg).gt(0)
# print('sum neg {} size {}'.format(torch.sum(neg),neg.size()))
# print('sum pos {}'.format(torch.sum(pos)))
# print('sum pos_and_neg {}'.format(torch.sum(pos_and_neg)))
# print('preds shape {}'.format(conf_preds.size()))
preds = conf_preds[mask].view(-1,self.num_classes)
targets = conf_targets[pos_and_neg]
conf_loss = F.cross_entropy(preds,targets,size_average=False)
N = num_pos.data.sum()
loc_loss /= N
conf_loss /= N
print('loc_loss:%f conf_loss:%f, pos_num:%d' % (loc_loss.data[0], conf_loss.data[0], N))
return loc_loss+conf_loss