-
Notifications
You must be signed in to change notification settings - Fork 0
/
p2ch14_malben_baseline.py
151 lines (123 loc) · 4.5 KB
/
p2ch14_malben_baseline.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
# ---
# jupyter:
# jupytext:
# text_representation:
# extension: .py
# format_name: light
# format_version: '1.5'
# jupytext_version: 1.16.3
# kernelspec:
# display_name: Python 3
# language: python
# name: python3
# ---
# +
import torch
# %matplotlib inline
from matplotlib import pyplot
import p2ch14.dsets
import p2ch14.model
# -
ds = p2ch14.dsets.MalignantLunaDataset(val_stride=10, isValSet_bool=True) # <1>
nodules = ds.ben_list + ds.mal_list
is_mal = torch.tensor([n.isMal_bool for n in nodules]) # <2>
diam = torch.tensor([n.diameter_mm for n in nodules])
num_mal = is_mal.sum() # <3>
num_ben = len(is_mal) - num_mal
threshold = torch.linspace(diam.max(), diam.min())
predictions = (diam[None] >= threshold[:, None]) # <1>
tp_diam = (predictions & is_mal[None]).sum(1).float() / num_mal # <2>
fp_diam = (predictions & ~is_mal[None]).sum(1).float() / num_ben
fp_diam_diff = fp_diam[1:] - fp_diam[:-1]
tp_diam_avg = (tp_diam[1:] + tp_diam[:-1])/2
auc_diam = (fp_diam_diff * tp_diam_avg).sum()
# +
fp_fill = torch.ones((fp_diam.shape[0] + 1,))
fp_fill[:-1] = fp_diam
tp_fill = torch.zeros((tp_diam.shape[0] + 1,))
tp_fill[:-1] = tp_diam
print(threshold)
print(fp_diam)
print(tp_diam)
# -
for i in range(threshold.shape[0]):
print(i, threshold[i], fp_diam[i], tp_diam[i])
pyplot.figure(figsize=(7,5), dpi=1200)
for i in [62, 88]:
pyplot.scatter(fp_diam[i], tp_diam[i], color='red')
print(f'diam: {round(threshold[i].item(), 2)}, x: {round(fp_diam[i].item(), 2)}, y: {round(tp_diam[i].item(), 2)}')
pyplot.fill(fp_fill, tp_fill, facecolor='#0077bb', alpha=0.25)
pyplot.plot(fp_diam, tp_diam, label=f'diameter baseline, AUC={auc_diam:.3f}')
pyplot.title(f'ROC diameter baseline, AUC={auc_diam:.3f}')
pyplot.ylabel('true positive rate')
pyplot.xlabel('false positive rate')
pyplot.savefig('roc_diameter_baseline.png')
model = p2ch14.model.LunaModel()
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
sd = torch.load('data/part2/models/cls_2020-02-08_01.19.40_finetune-head.best.state')
model.load_state_dict(sd['model_state'])
model.to(device)
model.eval();
ds = p2ch14.dsets.MalignantLunaDataset(val_stride=10, isValSet_bool=True)
dl = torch.utils.data.DataLoader(ds, batch_size=32, num_workers=4)
preds = []
truth = []
for inp, label, _, _, _ in dl:
inp = inp.to(device)
truth += (label[:,1]>0).tolist()
with torch.no_grad():
_, p = model(inp)
preds += p[:, 1].tolist()
truth = torch.tensor(truth)
preds = torch.tensor(preds)
# +
num_mal = truth.sum()
num_ben = len(truth) - num_mal
threshold = torch.linspace(1, 0)
tp_finetune = ((preds[None] >= threshold[:, None]) & truth[None]).sum(1).float() / num_mal
fp_finetune = ((preds[None] >= threshold[:, None]) & ~truth[None]).sum(1).float() / num_ben
fp_finetune_diff = fp_finetune[1:]-fp_finetune[:-1]
tp_finetune_avg = (tp_finetune[1:]+tp_finetune[:-1])/2
auc_finetune = (fp_finetune_diff * tp_finetune_avg).sum()
pyplot.figure(figsize=(7,5), dpi=300)
pyplot.fill(fp_fill, tp_fill, facecolor='#0077bb', alpha=0.25)
pyplot.plot(fp_diam, tp_diam, label=f'diameter baseline, AUC={auc_diam:.3f}')
pyplot.plot(fp_finetune, tp_finetune, label=f'1 layer fine-tuned, AUC={auc_finetune:.3f}')
pyplot.legend()
pyplot.savefig('roc_finetune.png')
# -
if 1:
fn = 'data/part2/models/cls_2020-02-08_00.19.45_finetune-depth2.best.state'
model = p2ch14.model.LunaModel()
sd = torch.load(fn, map_location='cpu')['model_state']
model.load_state_dict(sd)
model.to(device)
model.eval();
model.eval()
preds = []
truth = []
for inp, label, _, _, _ in dl:
inp = inp.to(device)
truth += (label[:,1]>0).tolist()
with torch.no_grad():
_, p = model(inp)
preds += p[:, 1].tolist()
truth = torch.tensor(truth)
preds = torch.tensor(preds)
# +
num_mal = truth.sum()
num_ben = len(truth) - num_mal
threshold = torch.linspace(1, 0)
tp = ((preds[None] >= threshold[:, None]) & truth[None]).sum(1).float() / num_mal
fp = ((preds[None] >= threshold[:, None]) & ~truth[None]).sum(1).float() / num_ben
fp_diff = fp[1:]-fp[:-1]
tp_avg = (tp[1:]+tp[:-1])/2
auc_modified = (fp_diff * tp_avg).sum()
pyplot.figure(figsize=(7,5), dpi=300)
pyplot.fill(fp_fill, tp_fill, facecolor='#0077bb', alpha=0.25)
pyplot.plot(fp_diam, tp_diam, label=f'diameter baseline, AUC={auc_diam:.3f}')
pyplot.plot(fp_finetune, tp_finetune, label=f'1 layer fine-tuned, AUC={auc_finetune:.3f}')
pyplot.plot(fp, tp, label=f'2 layers fine-tuned, AUC={auc_modified:.3f}')
pyplot.legend()
pyplot.savefig('roc_modified.png')
# -