-
Notifications
You must be signed in to change notification settings - Fork 0
/
EvaluateDir.py
165 lines (126 loc) · 4.95 KB
/
EvaluateDir.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Tue Oct 31 15:32:41 2017
@author: nigno
"""
# License: BSD
# Author: Sasank Chilamkurthy
from __future__ import print_function, division
from GazeNet import GazeNetRegDir
from MyLoss import LogLikeLoss
from DataLoadingDir import MirkoDatasetRegNorm, MirkoDatasetRegNormRam, Rescale, ToTensor, Normalize, RandomNoise
import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
from torch.autograd import Variable
import numpy as np
from torch.utils.data import Dataset
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
plt.ion() # interactive mode
def mse_loss(input, target):
return torch.sum((input - target).pow(2)) / input.data.nelement()
# Data augmentation and normalization for training
# Just normalization for validation
use_gpu = torch.cuda.is_available()
if use_gpu:
torchType = torch.cuda.FloatTensor
else:
torchType = torch.FloatTensor
data_transforms_custom1 = transforms.Compose([Rescale((240, 320)),
#RandomNoise(var = 0.02),
#Normalize(mean, std),
ToTensor()])
data_transforms_custom2 = transforms.Compose([Rescale((240, 320)),
#Normalize(mean, std),
ToTensor()])
seed = 1
root_dataset_test = 'Datasets/TestDatasetSimDirections/'
net_dir = 'Nets/alex_directions_full/'
checkpoint_dir = net_dir + 'checkpoints/'
model = GazeNetRegDir(1024)
if (use_gpu):
model = model.cuda()
print(model)
optimizer_ft = optim.Adam(model.parameters())
#checkpoint_name = 'checkpoint1530.tar'
checkpoint_name = 'checkpointAllEpochs.tar'
if os.path.isfile(checkpoint_dir + checkpoint_name):
print("=> loading checkpoint '{}'".format(checkpoint_dir + checkpoint_name))
checkpoint = torch.load(checkpoint_dir + checkpoint_name)
start_epoch = checkpoint['epoch']
best_abs = checkpoint['best_abs']
loss_list = checkpoint['loss_list']
abs_list = checkpoint['abs_list']
loss_list_val = checkpoint['loss_list_val']
abs_list_val = checkpoint['abs_list_val']
model.load_state_dict(checkpoint['state_dict'])
optimizer_ft.load_state_dict(checkpoint['optimizer'])
print("=> loaded checkpoint (epoch {})"
.format(checkpoint['epoch']))
model.train(False)
count = 0
time_elapsed_single = 0
n_folders = 8
# Generating the folders for the test set
test_root_dir_list = []
test_per_lis = []
for i in range(n_folders):
test_root_dir_list.append(root_dataset_test + str(i) + '/')
test_per_lis.append(0.0)
dataset_test = MirkoDatasetRegNorm(root_dir = test_root_dir_list,
transform = data_transforms_custom2,
dset_type='val', seed=seed,
training_per = test_per_lis,
permuted = False)
dataloader_test = torch.utils.data.DataLoader(dataset_test, batch_size=2,
shuffle=False, num_workers=8)
for data in dataloader_test:
# Go through a bunch of examples and record which are correctly guessed
start_time = time.time()
#data = dataset_test[count]
samples = data
# wrap them in Variable
if (use_gpu):
img = Variable(samples['image'].cuda())
label = Variable(samples['label'].cuda())
else:
img = Variable(samples['image'])
label = Variable(samples['label'])
start_time = time.time()
#PROVA-----------------------
#pred = model(img1.unsqueeze(0), img2.unsqueeze(0))
pred = model(img)
if (use_gpu):
pred_np = pred.cpu().data.numpy()
ground_truth_np = label.cpu().data.numpy()
else:
pred_np = pred.data.numpy()
ground_truth_np = label.data.numpy()
if count == 0:
predictions = pred_np
ground_truth = ground_truth_np
else:
predictions = np.concatenate((predictions, pred_np), axis=0)
ground_truth = np.concatenate((ground_truth, ground_truth_np), axis=0)
time_elapsed_single += time.time() - start_time
count += 1
time_elapsed_single /= count
print('Prediction complete in {:.5f}s'.format(time_elapsed_single))
print(predictions)
np.savetxt(net_dir + 'predictions.txt', predictions, delimiter=',')
np.savetxt(net_dir + 'ground_truth.txt', ground_truth, delimiter=',')
np.savetxt(net_dir + 'abs_list.txt', abs_list, delimiter=',')
np.savetxt(net_dir + 'abs_list_val.txt', abs_list_val, delimiter=',')
fig = plt.figure()
plt.plot(loss_list, color='red')
plt.plot(loss_list_val, color='blue')
fig = plt.figure()
plt.plot(abs_list, color='red')
plt.plot(abs_list_val, color='blue')
raw_input('Press enter to continue: ')