-
Notifications
You must be signed in to change notification settings - Fork 1
/
eval_map.py
52 lines (39 loc) · 1.28 KB
/
eval_map.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
import os
import numpy as np
import h5py
from average_precision import ap
import argparse
num_classes=21
label_file = 'data/test_frame.txt'
pred_path='prediction.h5'
parser = argparse.ArgumentParser(description='prediction file path.')
parser.add_argument('--pred_path', type=str, default='prediction.h5',
help='prediction file path')
args = parser.parse_args()
pred_path=args.pred_path
print(pred_path)
f = h5py.File(pred_path,'r')
processed_video=[]
preds =[]
labels =[]
with open(label_file, "r") as fp:
for string in fp:
string = string.split()
if(string[0] in processed_video):
continue
processed_video.append(string[0])
pred = f[string[0]+'/pred']
label = f[string[0]+'/label']
pred_reshape = np.zeros((pred.shape[0],pred.shape[1]), dtype=np.float32)
label_bool = np.zeros((pred.shape[0],pred.shape[1]), dtype=np.bool)
pred_reshape[:, 0:pred.shape[1]] = pred[:,:]
label_bool[:, 0:pred.shape[1]] = np.greater(label, 0.5)
labels.append(label_bool)
preds.append(pred_reshape)
preds = np.vstack(preds)
labels = np.vstack(labels)
ap = ap(labels,preds)
print('Average Precisions:')
print(ap)
print('mAP:')
print(np.average(ap[1:]))