forked from abdalazizrashid/idao-21-baseline
-
Notifications
You must be signed in to change notification settings - Fork 0
/
report.py
184 lines (153 loc) · 6.18 KB
/
report.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
182
183
184
import configparser
import gc
import logging
import pathlib as path
import sys
from collections import defaultdict
from itertools import chain
import matplotlib.pyplot as plt
import numpy as np
import scikitplot as skplt
import torch
from more_itertools import bucket
from idao.data_module import IDAODataModule
from idao.model import SimpleConv
from idao.utils import delong_roc_variance
def test_variance(target, predictions):
return torch.std(torch.abs(predictions - target) / target) ** 2
def run_test(mode, dataloader, checkpoint_path, cfg):
torch.multiprocessing.set_sharing_strategy("file_system")
logging.info("Loading checkpoint")
model = SimpleConv.load_from_checkpoint(checkpoint_path, mode=mode)
model = model.cpu().eval()
regression_predictions = []
classification_predictions = []
classification_target = []
regression_target = []
if mode == "classification":
logging.info("Classification model loaded")
else:
logging.info("Regression model loaded")
for i, (img, class_label, regression_label, _) in enumerate(iter(dataloader)):
if mode == "classification":
output = model(img)["class"].detach()
classification_predictions.append(output)
classification_target.append(class_label)
else:
output = model(img)["energy"].detach()
regression_predictions.append(output)
regression_target.append(regression_label)
# del output
if mode == "classification":
logging.info("Starting classification task")
classification_predictions = torch.cat(classification_predictions, dim=0)
classification_target = torch.cat(classification_target, dim=0)
new_target = np.argmax(classification_target.detach().cpu().numpy(), axis=1)
auc, variance = delong_roc_variance(
new_target, classification_predictions.detach().cpu().numpy()[:, 1]
)
skplt.metrics.plot_roc(
classification_target.max(1).indices,
classification_predictions.detach().cpu().numpy(),
plot_macro=False,
plot_micro=False,
classes_to_plot=[0]
)
plt.savefig(f'{cfg["REPORT"]["SaveDir"]}/roc_auc.png', dpi=196)
plt.show()
logging.info(f'ROC plot saved at: {cfg["REPORT"]["SaveDir"]}/roc_auc.png')
logging.info(f"Delong => ROC-AUC: {auc} variance: {variance}")
del classification_predictions
del classification_target
# gc.collect() # Invoke the garbage collector
return (None, auc)
else:
logging.info("Starting regression task")
regression_predictions = torch.tensor(
list(chain(*regression_predictions))
).view(-1)
regression_target = torch.tensor(list(chain(*regression_target))).view(-1)
variance = test_variance(regression_target, regression_predictions)
logging.info(f"Test energy variance: {variance}")
# MAE
mae = torch.nn.functional.l1_loss(regression_predictions, regression_target)
# plot correlation
fig, ax = plt.subplots()
ax.plot(
regression_target, regression_predictions, "ro", label="Energy Prediction"
)
ax.set_xlabel("True Energy")
ax.set_ylabel("Predicted Energy")
ax.legend()
ax.grid()
fig.savefig(f'{cfg["REPORT"]["SaveDir"]}/energy_correlation.png', dpi=196)
logging.info(f'Energy correlation plot saved at: {cfg["REPORT"]["SaveDir"]}/energy_correlation.png')
plt.close(fig)
logging.info(f'===> Length: {len(regression_predictions)} {len(regression_target)}')
# plot comparison
fig1, ax1 = plt.subplots()
ax1.plot(regression_predictions, "bo", alpha=0.6, label="Predicted Energy")
ax1.plot(regression_target, "ro", label="True Energy")
ax1.legend()
ax1.grid()
fig1.savefig(f'{cfg["REPORT"]["SaveDir"]}/energy_comparison.png', dpi=196)
logging.info(f'Energy comparison plot saved at: {cfg["REPORT"]["SaveDir"]}/energy_comparison.png')
plt.close(fig1)
# plot histograms
group = zip(regression_target, regression_predictions)
group = sorted(group, key=lambda item: item[0])
logging.info(regression_target)
data_dict = defaultdict(list)
for t, p in group:
data_dict[t.item()].append(p.item())
for i, (k, v) in enumerate(data_dict.items()):
fig3, ax3 = plt.subplots()
ax3.hist(
v,
bins=100,
histtype="step",
label=f"Energy: {k} keV \n RMS: {float(torch.sqrt(torch.mean(torch.tensor(v)**2))):.03f} \n Mean: {float(torch.mean(torch.tensor(v))):.03f}",
)
ax3.legend()
fig3.savefig(f'{cfg["REPORT"]["SaveDir"]}/energy_hist{k}_{i}.png', dpi=196)
plt.close(fig3)
logging.info(
f'Histogram {k} keV saved at: {cfg["REPORT"]["SaveDir"]}/energy_hist{k}_{i}.png'
)
return (mae, None)
def main(cfg):
PATH = path.Path(cfg["DATA"]["DatasetPath"])
dataset_dm = IDAODataModule(
data_dir=PATH, batch_size=64, cfg=cfg
)
dataset_dm.prepare_data()
dataset_dm.setup()
#dl = dataset_dm.test_dataloader()
dl = dataset_dm.train_dataloader()
mae = 0
variance = 0
for mode in ["regression", "classification"]:
if mode == "classification":
model_path = cfg["REPORT"]["ClassificationCheckpoint"]
else:
model_path = cfg["REPORT"]["RegressionCheckpoint"]
_mae, _auc = run_test(mode, dl, model_path, cfg=cfg)
if _mae is not None:
mae = _mae
if _auc is not None:
auc = _auc
gc.collect()
logging.info(f'MAE = {mae}')
logging.info(f'AUC = {auc}')
logging.info(f'Score = AUC - MAE: {auc - mae}')
if __name__ == "__main__":
config = configparser.ConfigParser()
config.read("./config.ini")
logging.basicConfig(
level=logging.INFO,
handlers=[
logging.FileHandler(f'{config["REPORT"]["SaveDir"]}report.log'),
logging.StreamHandler(sys.stdout),
],
)
main(cfg=config)