-
Notifications
You must be signed in to change notification settings - Fork 0
/
knn.py
96 lines (77 loc) · 3.12 KB
/
knn.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
from dataset import IndoorSceneFeatureDataset
from torch.utils.data import DataLoader
from sklearn.neighbors import KNeighborsClassifier
from utils import results
import torch
import time
def run(params):
indoorscene_traindataset = IndoorSceneFeatureDataset(
text_file='Dataset/TrainImages.txt',
feature_file=f'Dataset/{params["feature_extractor"]}-features.h5',
train=True)
train_loader = DataLoader(indoorscene_traindataset, batch_size=len(indoorscene_traindataset), shuffle=True,
num_workers=1)
indoorscene_testdataset = IndoorSceneFeatureDataset(
text_file='Dataset/TestImages.txt',
feature_file=f'Dataset/{params["feature_extractor"]}-features.h5',
train=False)
val_loader = DataLoader(indoorscene_testdataset, batch_size=len(indoorscene_testdataset), shuffle=True,
num_workers=1)
# training images
train_images, train_labels = next(iter(train_loader))
classes = indoorscene_traindataset.mapping
# training
clf = KNeighborsClassifier(
n_neighbors=len(classes),
algorithm=params['algorithm'],
weights=params['weight'],
n_jobs=16
)
train_start_time = time.time()
clf = clf.fit(train_images, train_labels)
train_time = time.time() - train_start_time
classes = indoorscene_traindataset.mapping
# train metrics
train_pred_start_time = time.time()
x_pred = clf.predict(train_images)
train_pred_time = time.time() - train_pred_start_time
x_pred_ohe = torch.zeros((len(x_pred), len(classes)))
for idx, lbl in enumerate(x_pred):
x_pred_ohe[idx][lbl] = 1
# test metrics
test_images, y_true = next(iter(val_loader))
test_pred_start_time = time.time()
y_pred = clf.predict(test_images)
test_pred_time = time.time() - test_pred_start_time
y_pred_ohe = torch.zeros((len(y_pred), len(classes)))
for idx, lbl in enumerate(y_pred):
y_pred_ohe[idx][lbl] = 1
# Saving metrics
params['train_time'] = train_time
params['train_pred_time'] = train_pred_time
params['test_pred_time'] = test_pred_time
# Store the results
results(train_labels, x_pred_ohe, y_true, y_pred_ohe, classes, params)
def run_loop():
type = 'knn'
feature_extractors = ['resnext101', 'mnasnet1_0']
algorithms = ['ball_tree','kd_tree','brute']
weights = ['uniform', 'distance']
count = 0
for feature_extractor in feature_extractors:
for algorithm in algorithms:
for weight in weights:
expt_name = f'{feature_extractor}-{algorithm}-{weight}'
total_experiments = len(feature_extractors) * len(algorithms) * \
len(weights)
print(f'{count}/{total_experiments}', type, expt_name)
run({
'weight':weight,
'algorithm':algorithm,
'feature_extractor':feature_extractor,
'exp_name' : expt_name,
'model_type':type
})
count += 1
if __name__ == '__main__':
run_loop()