-
Notifications
You must be signed in to change notification settings - Fork 3
/
test.py
132 lines (97 loc) · 4.31 KB
/
test.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
import os
import time
import torch
from utils import detect
from evaluation.evaluator import Evaluator
from config import device, device_ids
def test(test_loader, model, criterion, coder, opts):
# ---------- load ----------
model.eval()
state_dict = torch.load(os.path.join(opts.save_path, opts.save_file_name),
map_location=device)
model.load_state_dict(state_dict, strict=True)
tic = time.time()
sum_loss = 0
print('SKU110K dataset evaluation...')
evaluator = Evaluator(opts)
with torch.no_grad():
for idx, data in enumerate(test_loader):
images = data[0]
boxes = data[1]
labels = data[2]
locations = data[3]
counts = data[4]
map = data[5]
h = images.size(2)
w = images.size(3)
size = (h, w)
# ---------- cuda ----------
images = images.to(device)
boxes = [b.to(device) for b in boxes]
labels = [l.to(device) for l in labels]
locations = [(loc / opts.scale).to(device) for loc in locations]
counts = counts.to(device)
gt_map = map.to(device)
# ---------- loss ----------
pred = model(images)
loss, (cls_loss, loc_loss, obj_loss, cnt_loss) = criterion(pred, boxes, labels, locations, counts, gt_map, size)
sum_loss += loss.item()
# ---------- eval ----------
pred_boxes, pred_labels, pred_scores = detect(pred=pred[:2],
coder=coder,
opts=opts)
if opts.data_type == 'sku':
img_name = data[6][0]
img_info = data[7][0]
info = (pred_boxes, pred_labels, pred_scores, img_name, img_info)
evaluator.get_info(info)
toc = time.time()
# ---------- print ----------
if idx % opts.vis_step == 0 or idx == len(test_loader) - 1:
print('Step: [{0}/{1}]\t'
'Loss: {loss:.4f}\t'
'Time : {time:.4f}\t'
.format(idx, len(test_loader),
loss=loss,
time=toc - tic))
results = evaluator.evaluate(test_loader.dataset)
mAP = results[0]
mean_loss = sum_loss / len(test_loader)
# print("Avg Loss : ", mean_loss)
print("mAP : ", mAP)
print("Eval Time : {:.4f}".format(time.time() - tic))
if __name__ == "__main__":
from dataset.sku110_dataset import SKU110K_Dataset
from loss import IntegratedLoss
from model import HNNA_DET
from coder import RETINA_Coder
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('--save_path', type=str, default='./saves')
parser.add_argument('--save_file_name', type=str, default='pretrained_model.pth.tar')
parser.add_argument('--conf_thres', type=float, default=0.05)
parser.add_argument('--data_root', type=str, default='D:\SKU110K_fixed')
parser.add_argument('--data_type', type=str, default='sku')
parser.add_argument('--scale', type=int, default=8, help='image reduction scale')
parser.add_argument('--vis_step', type=int, default=100, help='image reduction scale')
parser.add_argument('--resize', type=int, default=800, help='image_size')
parser.add_argument('--num_classes', type=int, default=1)
test_opts = parser.parse_args()
print(test_opts)
vis = None
test_set = SKU110K_Dataset(root=test_opts.data_root, split='test')
test_opts.num_classes = 1
test_loader = torch.utils.data.DataLoader(test_set,
batch_size=1,
collate_fn=test_set.collate_fn,
shuffle=False,
num_workers=0)
model = HNNA_DET(num_classes=test_opts.num_classes).to(device)
model = torch.nn.DataParallel(module=model, device_ids=device_ids)
coder = RETINA_Coder(opts=test_opts)
criterion = IntegratedLoss(coder=coder)
test(test_loader=test_loader,
model=model,
criterion=criterion,
coder=coder,
opts=test_opts)