-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest_gaborpatches.py
181 lines (142 loc) · 7.55 KB
/
test_gaborpatches.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
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
from __future__ import print_function, division
# base libraries
import os
import glob
import argparse
from datetime import datetime
import csv
# libraries for arithmetic
import matplotlib
import matplotlib.pyplot as plt
import pandas as pd
# libraries for pytorch
import torch
import torch.nn as nn
from torchvision import transforms
# project based libraries
from utils import load_checkpoint
from dataloader import image_patch_loader, Quadrant_Processing
from models.VGG_Model import VGG, VGG_Freeze_conv, VGG_BN_Freeze_conv, VGG_Freeze_conv_FC1, VGG_Freeze_conv_FC2, Visual_Cortex_Amygdala, Visual_Cortex_Amygdala_wo_Attention
# cuda libraries
from GPUtil import showUtilization as gpu_usage
from numba import cuda
# import time
matplotlib.use('Agg')
plt.ion()
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = "3"
# Params
parser = argparse.ArgumentParser(description='Parameters ')
parser.add_argument('--data_dir', default='./data/gabor_RSA3', type=str, help='the data root folder')
parser.add_argument('--gabor_patch', default='gabor-gaussian-45-freq20-cont100.png', type=str, help='the folder of test data')
parser.add_argument('--model_to_run', default=1, type=int, help='which model you want to run with experiment')
parser.add_argument('--model_dir',
default='./savedmodel/Conditioning_Experiment13',
type=str, help='where to save the trained model')
parser.add_argument('--initial_model_name',
default='base_model.pth',
type=str, help='name of the trained model；'
'***Note**: You have to select which model can perform best on other datasets such as IAPS rather than the best model tested on the training data'
'Usually, the model that has the best generalizability is the trained model obtained at earlier epochs such as 3rd epoch.')
parser.add_argument('--conditioned_model_name',
default='base_model_epoch100.pth',
type=str, help='name of the trained model；'
'***Note**: You have to select which model can perform best on other datasets such as IAPS rather than the best model tested on the training data'
'Usually, the model that has the best generalizability is the trained model obtained at earlier epochs such as 3rd epoch.')
parser.add_argument('--gpu_ids', type=str, default='3', help='gpu ids: e.g. 0 0,1,2, 0,2. use -1 for CPU')
parser.add_argument('--csv_result_recording_folder', default='./result/model_test/', type=str,
help='the folder including the results csv')
parser.add_argument('--csv_result_recording_name', default='model_regression_test_results.csv')
'''
This code is used to determine if the model is conditioned or not by using gabor patch.
'''
def iter_run_allmodels(dir, modelName):
path = os.path.join(dir, modelName[:-4] + '*.pth')
for file_name in glob.iglob(path, recursive=True):
print(file_name)
def free_gpu_cache():
print("Initial GPU Usage")
gpu_usage()
torch.cuda.empty_cache()
cuda.select_device(0)
cuda.close()
cuda.select_device(0)
print("GPU Usage after emptying the cache")
gpu_usage()
if __name__ == '__main__':
# initialize
free_gpu_cache()
args = parser.parse_args()
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
initial_path = os.path.join(args.model_dir, args.initial_model_name[:-4] + '.pth')
conditioned_path = os.path.join(args.model_dir, args.conditioned_model_name[:-4] + '.pth')
file_list = [initial_path, conditioned_path]
# prepare the list of gabor patches to iterate.
gabor_list = []
for gabor_patch in os.listdir(args.data_dir):
if not gabor_patch.startswith('.'):
gabor_list.append(gabor_patch)
if len(file_list) < 1:
print('Cannot find models!')
else:
if not os.path.exists(args.csv_result_recording_folder):
os.makedirs(args.csv_result_recording_folder)
# variable initialization
initial_pred = 0
conditioned_pred = 0
result_df = pd.DataFrame(columns=['model', 'gabor', 'valence'])
# loop over all the files
for file_name in sorted(file_list):
print('\n Test model:{}'.format(file_name))
# Choose the model for analysis
if args.model_to_run == 1:
model = VGG()
elif args.model_to_run == 2:
model = VGG_Freeze_conv()
elif args.model_to_run == 3:
model = VGG_BN_Freeze_conv()
elif args.model_to_run == 4:
model = VGG_Freeze_conv_FC1()
elif args.model_to_run == 5:
model = VGG_Freeze_conv_FC2()
elif args.model_to_run == 6:
model = Visual_Cortex_Amygdala()
elif args.model_to_run == 7:
model = Visual_Cortex_Amygdala_wo_Attention()
# image size should be fixed as default as well as normalization parameters
image_size = 224
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
# defining transform for input gabor
gabor_patch_transform = transforms.Compose([
transforms.Resize(size=(image_size, image_size)),
transforms.CenterCrop(image_size),
Quadrant_Processing(2),
transforms.ToTensor(),
normalize])
# loop over all gabor patches
for gabor in sorted(gabor_list):
gabor_path = os.path.join(args.data_dir, gabor)
inputs = image_patch_loader(gabor_patch_transform, gabor_path)
# Load the model
model = load_checkpoint(model, file_name)
model = model.to(device)
inputs = inputs.to(device)
outputs = model(inputs)
outputs = 1 + outputs * (9 - 1) # normalize to 1-9 Because our network last layer is sigmoid,
# which gives 0-1 values to restrict the range of outputs to be [0-1] or [1-9]
# reshape the output to fit into dataframe.
preds = outputs.reshape(torch.Size([1])).cpu().detach().numpy()
if file_name == initial_path:
initial_pred = preds
print("The gabor patch used in test:", gabor)
print("The initial valence: {:.4f}".format(initial_pred[0]))
sub_df = pd.DataFrame([[args.initial_model_name, gabor, initial_pred[0]]], columns=['model', 'gabor', 'valence'])
result_df = result_df.append(sub_df, ignore_index=True)
else:
conditioned_pred = preds
print("The gabor patch used in test:", gabor)
print('Valence after conditioning : {:.4f}'.format(conditioned_pred[0]))
sub_df = pd.DataFrame([[args.conditioned_model_name, gabor, conditioned_pred[0]]],columns=['model', 'gabor', 'valence'])
result_df = result_df.append(sub_df, ignore_index=True)
result_df.to_csv(args.csv_result_recording_folder + 'Conditioning_Experiment' + '_' + args.data_dir[24:] + '.csv')