-
Notifications
You must be signed in to change notification settings - Fork 30
/
val.py
97 lines (83 loc) · 2.77 KB
/
val.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
import os, json, faiss
from utils.utils import *
from tqdm import tqdm
from models.resnet import *
from config.config import Config
def get_vector2index():
return faiss.IndexFlatL2(512)
class CFG():
def __init__(self):
self.vector2index = get_vector2index()
def read_val(path_val, data_root):
files = open(path_val, 'r+')
dict_data = []
for line in files.read().splitlines():
if 'song' in line:
type = 'song'
else:
type = 'hum'
dict_data.append({
'path': os.path.join(data_root, line.split(' ')[0]),
'label': line.split(' ')[1],
'type': type
})
files.close()
return dict_data
def search_vector(path_hum, model, index2id, cfg, input_shape):
image = load_image(path_hum, input_shape)
feature = get_feature(model, image)
distances, lst_index = cfg.vector2index.search(feature, k=10)
lst_result = []
for index in lst_index[0]:
result = str(index2id[str(index)])
lst_result.append(result)
return lst_result, distances
def mean_reciprocal_rank(preds, gt: str, k: int=10):
preds = preds[: min(k, len(preds))]
score = 0
for rank, pred in enumerate(preds):
if pred == gt:
score = 1 / (rank + 1)
break
return score
def mrr_score(model, dict_data, input_shape):
index2id = {"-1": ""}
id = 0
cfg = CFG()
count = 0
s_0 = 0
s_1 = 0
result_search = []
for item in tqdm(dict_data):
if item['type'] == 'song':
path_song = item['path']
image = load_image(path_song, input_shape)
cfg.vector2index.add(get_feature(model, image))
index2id[str(id)] = item['label']
id += 1
for item in tqdm(dict_data):
if item['type'] == 'hum':
path_hum = item['path']
preds, distances = search_vector(path_hum, model, index2id, cfg, input_shape)
result_search.append([item['label'], preds])
mrr = mean_reciprocal_rank(preds, item['label'])
if mrr == 1.0:
s_0 += distances[0, 0]
s_1 += distances[0, 1]
count += 1
mrr = 0
for row in result_search:
mrr += mean_reciprocal_rank(row[1], row[0])
mrr = mrr/len(result_search)
print("AVG", s_0 / count, s_1 / count, count)
return mrr
if __name__ == '__main__':
config = Config()
model = wrap_resnet_face18(config.use_se)
model.load_state_dict(torch.load(os.path.join(config.checkpoints_path, 'resnet18_latest.pth')))
model.to('cuda')
model.eval()
dict_data = read_val(config.val_list, config.train_root)
mrr = mrr_score(model, dict_data, config.input_shape)
print("-----------------------------")
print(f"MRR: {mrr}")