File size: 12,094 Bytes
f3972ea
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
from utils import CustomDataset, transform, Convert_ONNX
from torch.utils.data import Dataset, DataLoader
from utils import CustomDataset, TestingDataset, transform
from tqdm import tqdm
import torch
import numpy as np
from resnet_model_mask import  ResidualBlock, ResNet
import torch
import torch.nn as nn
import torch.optim as optim
from tqdm import tqdm 
import torch.nn.functional as F
from torch.optim.lr_scheduler import ReduceLROnPlateau
import pickle
import matplotlib.pyplot as plt
import pandas as pd

torch.manual_seed(1)
# torch.manual_seed(42)


device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
num_gpus = torch.cuda.device_count()
print(num_gpus)

test_data_dir = '/mnt/buf1/pma/frbnn/test_ready'
test_dataset = TestingDataset(test_data_dir, transform=transform)

num_classes = 2
testloader = DataLoader(test_dataset, batch_size=420, shuffle=True, num_workers=32)

model = ResNet(24, ResidualBlock, [3, 4, 6, 3], num_classes=num_classes).to(device)
model = nn.DataParallel(model)
model = model.to(device)
params = sum(p.numel() for p in model.parameters())
print("num params ",params)

model_1 = 'models_mask/model-43-99.235_42.pt'
# model_1 ='models/model-47-99.125.pt'
model.load_state_dict(torch.load(model_1, weights_only=True))
model = model.eval()

# eval
val_loss = 0.0
correct_valid = 0
total = 0
results = {'output': [],'pred': [], 'true':[], 'freq':[], 'snr':[], 'dm':[], 'boxcar':[]}
model.eval()
with torch.no_grad():
    for images, labels in tqdm(testloader):
        inputs, labels = images.to(device), labels
        outputs = model(inputs, return_mask = True)
        _, predicted = torch.max(outputs, 1)
        results['output'].extend(outputs.cpu().numpy().tolist())
        results['pred'].extend(predicted.cpu().numpy().tolist())
        results['true'].extend(labels[0].cpu().numpy().tolist())
        results['freq'].extend(labels[2].cpu().numpy().tolist())
        results['dm'].extend(labels[1].cpu().numpy().tolist())
        results['snr'].extend(labels[3].cpu().numpy().tolist())
        results['boxcar'].extend(labels[4].cpu().numpy().tolist())
        total += labels[0].size(0)
        correct_valid += (predicted.cpu() == labels[0].cpu()).sum().item()
    
# Calculate training accuracy after each epoch
val_accuracy = correct_valid / total * 100.0
print("===========================")
print('accuracy: ',  val_accuracy)
print("===========================")

import pickle

# Pickle the dictionary to a file
with open('models_mask/test_42.pkl', 'wb') as f:
    pickle.dump(results, f)

from sklearn.metrics import precision_score, recall_score, f1_score, confusion_matrix

# Example binary labels
true = results['true']  # ground truth
pred = results['pred']  # predicted

# Compute metrics
precision = precision_score(true, pred)
recall = recall_score(true, pred)
f1 = f1_score(true, pred)
# Get confusion matrix: TN, FP, FN, TP
tn, fp, fn, tp = confusion_matrix(true, pred).ravel()

# Compute FPR
fpr = fp / (fp + tn)

print(f"False Positive Rate: {fpr:.3f}")

print(f"Precision: {precision:.3f}")
print(f"Recall: {recall:.3f}")
print(f"F1 Score: {f1:.3f}")

# Plot accuracy as a function of DM
# Create a DataFrame for easier manipulation
df = pd.DataFrame({
    'dm': results['dm'],
    'true': results['true'],
    'pred': results['pred'],
    'snr': results['snr'],
    'freq': results['freq'],
    'boxcar': np.array(results['boxcar'])/2
})

# Filter to only include positive class samples (true == 1)
df = df[df['true'] == 1].copy()

print(f"Filtered to {len(df)} samples with true label = 1")

# Create DM bins for grouping
dm_bins = np.linspace(df['dm'].min(), df['dm'].max(), 20)  # 20 bins
df['dm_bin'] = pd.cut(df['dm'], bins=dm_bins, include_lowest=True)
print('min boxcar',df['boxcar'].min())
# Calculate accuracy and uncertainty for each DM bin
def calculate_accuracy_with_uncertainty(group):
    correct = (group['true'] == group['pred']).sum()
    total = len(group)
    accuracy = correct / total * 100
    # Standard error for binomial proportion
    p = correct / total
    se = np.sqrt(p * (1 - p) / total) * 100  # Convert to percentage
    return pd.Series({'accuracy': accuracy, 'std_error': se, 'n_samples': total})

dm_accuracy = df.groupby('dm_bin').apply(calculate_accuracy_with_uncertainty).reset_index()

# Get the midpoint of each bin for plotting
dm_accuracy['dm_midpoint'] = dm_accuracy['dm_bin'].apply(lambda x: x.mid)

# Create the plot with error bars
plt.figure(figsize=(10, 6))
ax1 = plt.gca()
ax1.errorbar(dm_accuracy['dm_midpoint'], dm_accuracy['accuracy'], 
             yerr=dm_accuracy['std_error'], fmt='o-', color='#b80707', linewidth=2, markersize=6,
             capsize=5, capthick=2, elinewidth=1)
ax1.set_xlabel('Dispersion Measure (DM) [pc cm$^{-3}$]', fontsize=16)
ax1.set_ylabel('Accuracy (%)', fontsize=16)
ax1.set_title('Accuracy vs Dispersion Measure', fontsize=18)
ax1.grid(True, alpha=0.3)
ax1.set_ylim(97, 100)
ax1.tick_params(axis='both', which='major', labelsize=14)

# Remove y-axis ticks over 100
yticks = ax1.get_yticks()
ax1.set_yticks([tick for tick in yticks if tick <= 100])

# Add some statistics to the plot
ax1.axhline(y=val_accuracy, color='r', linestyle='--', alpha=0.7, 
            label=f'Overall: {val_accuracy:.2f}%')
ax1.legend(fontsize=14)

# Add subplot labels at the bottom
ax1.text(-0.1, -0.15, '(a)', transform=ax1.transAxes, fontsize=18, fontweight='bold')

plt.tight_layout()
plt.savefig('models_mask/accuracy_vs_dm.pdf', dpi=300, bbox_inches='tight')
plt.show()

# Plot accuracy as a function of SNR
# Filter out zero/negative SNR values (not physically meaningful)
df_snr_filtered = df[df['snr'] > 0].copy()

# Create SNR bins for grouping
snr_bins = np.linspace(df_snr_filtered['snr'].min(), df_snr_filtered['snr'].max(), 20)  # 20 bins
df_snr_filtered['snr_bin'] = pd.cut(df_snr_filtered['snr'], bins=snr_bins, include_lowest=True)

# Calculate accuracy and uncertainty for each SNR bin
snr_accuracy = df_snr_filtered.groupby('snr_bin').apply(calculate_accuracy_with_uncertainty).reset_index()

# Get the midpoint of each bin for plotting
snr_accuracy['snr_midpoint'] = snr_accuracy['snr_bin'].apply(lambda x: x.mid)

# Create the SNR plot with error bars
plt.figure(figsize=(10, 6))
ax2 = plt.gca()
ax2.errorbar(snr_accuracy['snr_midpoint'], snr_accuracy['accuracy'], 
             yerr=snr_accuracy['std_error'], fmt='o-', color='#b80707', linewidth=2, markersize=6,
             capsize=5, capthick=2, elinewidth=1)
ax2.set_xlabel('Signal-to-Noise Ratio (SNR)', fontsize=16)
ax2.set_ylabel('Accuracy (%)', fontsize=16)
ax2.set_title('Accuracy vs SNR', fontsize=18)
ax2.grid(True, alpha=0.3)
ax2.set_ylim(80, 100)
ax2.tick_params(axis='both', which='major', labelsize=14)

# Remove y-axis ticks over 100
yticks = ax2.get_yticks()
ax2.set_yticks([tick for tick in yticks if tick <= 100])

# Add overall accuracy reference line
ax2.axhline(y=val_accuracy, color='r', linestyle='--', alpha=0.7, 
            label=f'Overall: {val_accuracy:.2f}%')
ax2.legend(fontsize=14)

# Add subplot labels at the bottom
ax2.text(-0.1, -0.15, '(b)', transform=ax2.transAxes, fontsize=18, fontweight='bold')

plt.tight_layout()
plt.savefig('models_mask/accuracy_vs_snr.pdf', dpi=300, bbox_inches='tight')
plt.show()

# Plot accuracy as a function of boxcar
# Create boxcar bins for grouping
# Use quantile-based binning to ensure all bins have data
# Filter out zero/negative values first for meaningful analysis
df_boxcar_filtered = df[df['boxcar'] > 0].copy()
df_boxcar_filtered['boxcar_bin'] = pd.qcut(df_boxcar_filtered['boxcar'], q=20, duplicates='drop')

# Calculate accuracy and uncertainty for each boxcar bin
boxcar_accuracy = df_boxcar_filtered.groupby('boxcar_bin').apply(calculate_accuracy_with_uncertainty).reset_index()

# Get the midpoint of each bin for plotting
boxcar_accuracy['boxcar_midpoint'] = boxcar_accuracy['boxcar_bin'].apply(lambda x: x.mid)

# Create the boxcar plot with error bars
plt.figure(figsize=(10, 6))
ax3 = plt.gca()
ax3.errorbar(boxcar_accuracy['boxcar_midpoint'], boxcar_accuracy['accuracy'], 
             yerr=boxcar_accuracy['std_error'], fmt='o-', color='#b80707', linewidth=2, markersize=6,
             capsize=5, capthick=2, elinewidth=1)
ax3.set_xscale('log')
ax3.set_xlabel('Boxcar Width (log scale)', fontsize=16)
# ax3.set_title('Accuracy vs Boxcar Width (Log Scale)', fontsize=18)
ax3.grid(True, alpha=0.3)
ax3.set_ylim(0, 100)
ax3.tick_params(axis='both', which='major', labelsize=14)

# Remove y-axis ticks over 100
yticks = ax3.get_yticks()
ax3.set_yticks([tick for tick in yticks if tick <= 100])

# Add overall accuracy reference line
ax3.axhline(y=val_accuracy, color='r', linestyle='--', alpha=0.7, 
            label=f'Overall: {val_accuracy:.2f}%')
ax3.legend(fontsize=14)

# Add subplot labels at the bottom
ax3.text(-0.1, -0.15, '(c)', transform=ax3.transAxes, fontsize=18, fontweight='bold')

plt.tight_layout()
plt.savefig('models_mask/accuracy_vs_boxcar.pdf', dpi=300, bbox_inches='tight')
plt.show()


print(f"Plots saved to models_mask/accuracy_vs_dm.pdf, models_mask/accuracy_vs_snr.pdf, and models_mask/accuracy_vs_boxcar.pdf")

# Create combined plot with all three parameters
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(18, 6))

# DM plot with error bars
ax1.errorbar(dm_accuracy['dm_midpoint'], dm_accuracy['accuracy'], 
             yerr=dm_accuracy['std_error'], fmt='o-', color='#b80707', linewidth=2, markersize=6,
             capsize=5, capthick=2, elinewidth=1)
ax1.set_xlabel('Dispersion Measure (DM) [pc cm$^{-3}$]', fontsize=16)
ax1.set_ylabel('Accuracy (%)', fontsize=16)
# ax1.set_title('Accuracy vs DM', fontsize=18)
ax1.grid(True, alpha=0.3)
ax1.set_ylim(97, 100.5)
ax1.tick_params(axis='both', which='major', labelsize=14)

# Remove y-axis ticks over 100
yticks = ax1.get_yticks()
ax1.set_yticks([tick for tick in yticks if tick <= 100])

ax1.axhline(y=val_accuracy, color='r', linestyle='--', alpha=0.7, 
            label=f'Overall: {val_accuracy:.2f}%')
ax1.legend(fontsize=14)

# SNR plot with error bars
ax2.errorbar(snr_accuracy['snr_midpoint'], snr_accuracy['accuracy'], 
             yerr=snr_accuracy['std_error'], fmt='o-', color='#b80707', linewidth=2, markersize=6,
             capsize=5, capthick=2, elinewidth=1)
ax2.set_xlabel('Signal-to-Noise Ratio (SNR)', fontsize=16)
# ax2.set_title('Accuracy vs SNR', fontsize=18)
ax2.grid(True, alpha=0.3)
ax2.set_ylim(88, 100.5)
ax2.tick_params(axis='both', which='major', labelsize=14)

# Remove y-axis ticks over 100
yticks = ax2.get_yticks()
ax2.set_yticks([tick for tick in yticks if tick <= 100])

ax2.axhline(y=val_accuracy, color='r', linestyle='--', alpha=0.7, 
            label=f'Overall: {val_accuracy:.2f}%')
ax2.legend(fontsize=14)

# Boxcar plot (log scale) with error bars
ax3.errorbar(boxcar_accuracy['boxcar_midpoint'][:-1],
             boxcar_accuracy['accuracy'][:-1], 
             yerr=boxcar_accuracy['std_error'][:-1], fmt='o-', color='#b80707', linewidth=2, markersize=6,
             capsize=5, capthick=2, elinewidth=1)
ax3.set_xscale('log')
ax3.set_xlabel('Boxcar Width (log scale) [s]', fontsize=16)
# ax3.set_title('Accuracy vs Boxcar Width', fontsize=18)
ax3.grid(True, alpha=0.3)
ax3.set_ylim(96, 100.5)
ax3.tick_params(axis='both', which='major', labelsize=14)

# Remove y-axis ticks over 100
yticks = ax3.get_yticks()
ax3.set_yticks([tick for tick in yticks if tick <= 100])

ax3.axhline(y=val_accuracy, color='r', linestyle='--', alpha=0.7, 
            label=f'Overall: {val_accuracy:.2f}%')
ax3.legend(fontsize=14)

# Add subplot labels at the bottom
ax1.text(-0.1, -0.15, '(a)', transform=ax1.transAxes, fontsize=18, fontweight='bold')
ax2.text(-0.1, -0.15, '(b)', transform=ax2.transAxes, fontsize=18, fontweight='bold')
ax3.text(-0.1, -0.15, '(c)', transform=ax3.transAxes, fontsize=18, fontweight='bold')

plt.tight_layout()
plt.savefig('models_mask/accuracy_vs_all_parameters.pdf', 
            dpi=300, bbox_inches='tight',
            pad_inches=0.1, format='pdf')
plt.show()

print(f"Combined plot saved to models_mask/accuracy_vs_all_parameters.pdf")