-
Notifications
You must be signed in to change notification settings - Fork 3
/
ensemble_ntu_cs.py
66 lines (57 loc) · 2.42 KB
/
ensemble_ntu_cs.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
import pickle
import numpy as np
from tqdm import tqdm
# Linear
print('-' * 20 + 'Linear Eval' + '-' * 20)
# joint_path = 'work_dir/SkeAttnCLR_/xsub/joint/linear/'
# bone_path = 'work_dir/SkeAttnCLR_/xsub/bone/linear/'
# motion_path = 'work_dir/SkeAttnCLR_/xsub/motion/linear/'
# joint_path = 'work_dir/SkeAttnCLR_/xsub/joint/semi0.1/'
# bone_path = 'work_dir/SkeAttnCLR_/xsub/bone/semi0.1/'
# motion_path = 'work_dir/SkeAttnCLR_/xsub/motion/semi0.1/'
joint_path = 'work_dir/SkeletonCLR_PKU/xsub/joint/linear/'
bone_path = 'work_dir/SkeletonCLR_PKU/xsub/bone/linear/'
motion_path = 'work_dir/SkeletonCLR_PKU/xsub/motion/linear/'
# joint_path = 'work_dir/SkeAttnCLR_/xsub/joint/finetune/'
# bone_path = 'work_dir/SkeAttnCLR_/xsub/bone/finetune/'
# motion_path = 'work_dir/SkeAttnCLR_/xsub/motion/finetune/'
# joint_path = 'work_dir/SkeAttnCLR_NTU120_/xsub/joint/finetune/'
# bone_path = 'work_dir/SkeAttnCLR_NTU120/xsub/bone/finetune/'
# motion_path = 'work_dir/SkeAttnCLR_NTU120/xsub/motion/finetune/'
#label = open('/mnt/netdisk/linlilang/CrosSCLR/data/NTU-RGB-D_120/xsub/val_label.pkl', 'rb')
# label = open('data/NTU60/xsub/val_label.pkl', 'rb')
# label = open('data/NTU120/NTU120_origin/xsub/val_label.pkl', 'rb')
# label = np.array(pickle.load(label))
label = open('data/PKU_part1_frame64/xsub/val_label.pkl', 'rb')
label = np.array(pickle.load(label))
r1 = open(joint_path + 'best_test_result.pkl', 'rb')
r1 = list(pickle.load(r1).items())
r2 = open(bone_path + 'best_test_result.pkl', 'rb')
r2 = list(pickle.load(r2).items())
r3 = open(motion_path + 'best_test_result.pkl', 'rb')
r3 = list(pickle.load(r3).items())
alpha1 = [0.5,0.5,0.5]# average fusion
alpha2 = [0.5, 0.5, 0.4] # aimclr weighted fusion
if 1:
best = 0.0
best_weight = []
for alpha in [alpha1,alpha2]:
right_num = total_num = right_num_5 = 0
for i in tqdm(range(len(label[0]))):
_, l = label[:, i]
_, r11 = r1[i]
_, r22 = r2[i]
_, r33 = r3[i]
r = r11 * alpha[0] + r22 * alpha[1] + r33 * alpha[2]
rank_5 = r.argsort()[-5:]
right_num_5 += int(int(l) in rank_5)
r = np.argmax(r)
right_num += int(r == int(l))
total_num += 1
acc = right_num / total_num
acc5 = right_num_5 / total_num
if acc> best:
best = acc
best_weight = alpha
print(alpha, 'top1: ', acc)
print(best_weight, best)