-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathtest_ICFG_my.py
170 lines (140 loc) · 5.74 KB
/
test_ICFG_my.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
from option.options import options, config
from data.dataloader import get_dataloader
import torch
import random
from model.model import TextImgPersonReidNet
from loss.Id_loss import Id_Loss
from loss.RankingLoss import RankingLoss
from torch import optim
import logging
import os
from test_during_train import test , test_part
from torch.autograd import Variable
from model.DETR_model import TextImgPersonReidNet_mydecoder_pixelVit_transTXT_3, TextImgPersonReidNet_mydecoder_pixelVit_transTXT_3_vit
import torch.nn as nn
logger = logging.getLogger()
logger.setLevel(logging.INFO)
def save_checkpoint(state, opt):
filename = os.path.join(opt.save_path, 'model/best.pth.tar')
torch.save(state, filename)
def load_checkpoint(opt):
filename = os.path.join(opt.save_path, 'model/best.pth.tar')
state = torch.load(filename)
return state
def calculate_similarity(image_embedding, text_embedding):
image_embedding_norm = image_embedding / image_embedding.norm(dim=1, keepdim=True)
text_embedding_norm = text_embedding / text_embedding.norm(dim=1, keepdim=True)
similarity = torch.mm(image_embedding_norm, text_embedding_norm.t())
return similarity
def calculate_similarity_part(numpart,image_embedding, text_embedding):
image_embedding = torch.cat([image_embedding[i] for i in range(numpart)],dim=1)
text_embedding = torch.cat([text_embedding[i] for i in range(numpart)], dim=1)
image_embedding_norm = image_embedding / image_embedding.norm(dim=1, keepdim=True)
text_embedding_norm = text_embedding / text_embedding.norm(dim=1, keepdim=True)
similarity = torch.mm(image_embedding_norm, text_embedding_norm.t())
return similarity
def calculate_similarity_score(opt,image_embedding, text_embedding , img_score ,txt_score):
img_size = img_score.size(0)
txt_size = txt_score.size(0)
part_num = img_score.size(1)
Final_matrix = torch.FloatTensor(img_size, txt_size).zero_().to(opt.device)
Fq_matrix = torch.FloatTensor(img_size, txt_size).zero_().to(opt.device)
for i in range(part_num):
# print(i)
# Compute pairwise distance, replace by the official when merged
image_embedding_i = image_embedding[i]
text_embedding_i = text_embedding[i]
image_embedding_i = image_embedding_i / image_embedding_i.norm(dim=1, keepdim=True)
text_embedding_i = text_embedding_i / text_embedding_i.norm(dim=1, keepdim=True)
similarity = torch.mm(image_embedding_i, text_embedding_i.t())
img_score_i = img_score[:, i].unsqueeze(1) # .view(q_score.size(0), 1)
txt_score_i = txt_score[:, i].unsqueeze(1)
# print(img_score.shape)
# print(img_score_i.shape)
q_matrix = torch.mm(img_score_i, txt_score_i.t())
final_matrix = similarity.mul(q_matrix)
Final_matrix = Final_matrix + final_matrix
Fq_matrix = Fq_matrix + q_matrix
Fq_matrix = Fq_matrix + 1e-12
# print(Fq_matrix)
dist_part = torch.div(Final_matrix, Fq_matrix)
return dist_part
if __name__ == '__main__':
opt = options().opt
opt.GPU_id = '1'
opt.device = torch.device('cuda:{}'.format(opt.GPU_id))
opt.data_augment = False
opt.lr = 0.001
opt.margin = 0.2
opt.feature_length = 512
opt.train_dataset = 'CUHK-PEDES'
opt.dataset = 'MSMT-PEDES'
if opt.dataset == 'MSMT-PEDES':
opt.pkl_root = '/data1/zhiying/text-image/MSMT-PEDES/3-1/'
opt.class_num = 3102
opt.vocab_size = 2500
opt.dataroot = '/data1/zhiying/text-image/data/ICFG_PEDES'
# opt.class_num = 2802
# opt.vocab_size = 2300
elif opt.dataset == 'CUHK-PEDES':
opt.pkl_root = '/data1/zhiying/text-image/CUHK-PEDES_/' # same_id_new_
opt.class_num = 11003
opt.vocab_size = 5000
opt.dataroot = '/data1/zhiying/text-image/CUHK-PEDES'
opt.d_model = 1024
opt.nhead = 4
opt.dim_feedforward = 2048
opt.normalize_before = False
opt.num_encoder_layers = 3
opt.num_decoder_layers = 3
opt.num_query = 6
opt.detr_lr = 0.0001
opt.txt_detr_lr = 0.0001
opt.txt_lstm_lr = 0.001
opt.res_y = False
opt.noself = False
opt.post_norm = False
opt.n_heads = 4
opt.n_layers = 2
opt.share_query = True
model_name = 'random_my_small_DeiT_2version_head6_384lstm'
# model_name = 'test'
opt.save_path = './checkpoints/dual_modal/{}/'.format(opt.train_dataset) + model_name
opt.epoch = 60
opt.epoch_decay = [20, 40, 50]
opt.batch_size = 64
opt.start_epoch = 0
opt.trained = False
config(opt)
opt.epoch_decay = [i - opt.start_epoch for i in opt.epoch_decay]
train_dataloader = get_dataloader(opt)
opt.mode = 'test'
test_img_dataloader, test_txt_dataloader = get_dataloader(opt)
opt.mode = 'train'
# train_dataloader = get_dataloader(opt)
# opt.mode = 'test'
# test_img_dataloader, test_txt_dataloader = get_dataloader(opt)
# opt.mode = 'train'
id_loss_fun = nn.ModuleList()
for _ in range(opt.num_query):
id_loss_fun.append(Id_Loss(opt).to(opt.device))
ranking_loss_fun = RankingLoss(opt)
network = TextImgPersonReidNet_mydecoder_pixelVit_transTXT_3_vit(opt).to(opt.device)
test_best = 0
test_history = 0
state = load_checkpoint(opt)
network.load_state_dict(state['network'])
test_best = state['test_best']
test_history = test_best
print('load the {} epoch param successfully'.format(state['epoch']))
"""
network.eval()
test_best = test(opt, 0, 0, network,
test_img_dataloader, test_txt_dataloader, test_best)
network.train()
exit(0)
"""
network.eval()
test_best = test_part(opt, state['epoch'], 1, network,
test_img_dataloader, test_txt_dataloader, test_best)
logging.info('Training Done')