-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathinference.py
165 lines (149 loc) · 7.48 KB
/
inference.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
import sys
import numpy as np
import pandas as pd
import os
import cv2
import wandb
from datetime import datetime
from tqdm import tqdm
import argparse
import random
import copy
import json
import csv
import torch
from torch.utils.data import DataLoader, Dataset, ConcatDataset
from sklearn.metrics import accuracy_score, f1_score
from hc701fed.dataset.val_dataset_list import (
APTOS_Val,
EyePACS_Val,
MESSIDOR_2_Val,
MESSIDOR_pairs_Val,
MESSIDOR_Etienne_Val,
MESSIDOR_Brest_Val,
)
from hc701fed.dataset.val_dataset_list import (
APTOS_Test,
EyePACS_Test,
MESSIDOR_2_Test,
MESSIDOR_pairs_Test,
MESSIDOR_Etienne_Test,
MESSIDOR_Brest_Test,
)
from hc701fed.model.baseline import Baseline
def main(backbone,model_path,dataset,trained_dataset,mode,device):
if model_path == "none":
raise NotImplementedError("Please specify the model path")
if mode == 'val':
if dataset == "centerlized":
Centerlized_Val = ConcatDataset([APTOS_Val, EyePACS_Val, MESSIDOR_2_Val, MESSIDOR_pairs_Val, MESSIDOR_Etienne_Val, MESSIDOR_Brest_Val])
val_dataset = DataLoader(Centerlized_Val, batch_size = 256, shuffle=False)
elif dataset == "messidor":
MESSIDOR_Centerlized_Val = ConcatDataset([MESSIDOR_pairs_Val, MESSIDOR_Etienne_Val, MESSIDOR_Brest_Val])
val_dataset = DataLoader(MESSIDOR_Centerlized_Val, batch_size = 256, shuffle=False)
elif dataset == "aptos":
val_dataset = DataLoader(APTOS_Val, batch_size = 256, shuffle=False)
elif dataset == "eyepacs":
val_dataset = DataLoader(EyePACS_Val, batch_size = 256, shuffle=False)
elif dataset == "messidor2":
val_dataset = DataLoader(MESSIDOR_2_Val, batch_size = 256, shuffle=False)
elif dataset == "messidor_pairs":
val_dataset = DataLoader(MESSIDOR_pairs_Val, batch_size = 256, shuffle=False)
elif dataset == "messidor_etienne":
val_dataset = DataLoader(MESSIDOR_Etienne_Val, batch_size = 256, shuffle=False)
elif dataset == "messidor_brest":
val_dataset = DataLoader(MESSIDOR_Brest_Val, batch_size = 256, shuffle=False)
else:
raise NotImplementedError
elif mode == 'test':
if dataset == "centerlized":
Centerlized_Test = ConcatDataset([APTOS_Test, EyePACS_Test, MESSIDOR_2_Test, MESSIDOR_pairs_Test, MESSIDOR_Etienne_Test, MESSIDOR_Brest_Test])
val_dataset = DataLoader(Centerlized_Test, batch_size = 256, shuffle=False)
elif dataset == "messidor":
MESSIDOR_Centerlized_Test = ConcatDataset([MESSIDOR_pairs_Test, MESSIDOR_Etienne_Test, MESSIDOR_Brest_Test])
val_dataset = DataLoader(MESSIDOR_Centerlized_Test, batch_size = 256, shuffle=False)
elif dataset == "aptos":
val_dataset = DataLoader(APTOS_Test, batch_size = 256, shuffle=False)
elif dataset == "eyepacs":
val_dataset = DataLoader(EyePACS_Test, batch_size = 256, shuffle=False)
elif dataset == "messidor2":
val_dataset = DataLoader(MESSIDOR_2_Test, batch_size = 256, shuffle=False)
elif dataset == "messidor_pairs":
val_dataset = DataLoader(MESSIDOR_pairs_Test, batch_size = 256, shuffle=False)
elif dataset == "messidor_etienne":
val_dataset = DataLoader(MESSIDOR_Etienne_Test, batch_size = 256, shuffle=False)
elif dataset == "messidor_brest":
val_dataset = DataLoader(MESSIDOR_Brest_Test, batch_size = 256, shuffle=False)
else:
raise NotImplementedError
else:
raise NotImplementedError
if trained_dataset == 'messidor' or trained_dataset == 'messidor_pairs' or trained_dataset == 'messidor_etienne' or trained_dataset == 'messidor_brest':
num_classes = 4
else:
num_classes = 5
model_load_path = os.path.join(model_path, f"{trained_dataset}_{backbone}_best.pth")
model = Baseline(backbone = backbone, num_classes = num_classes, pretrained = False)
model.load_state_dict(torch.load(model_load_path))
model.to(device)
model.eval()
y_true = []
y_pred = []
y_pred_prob = []
# print inference setting
print(f"Dataset: {dataset}, Backbone: {backbone}, Mode: {mode}, trained_dataset: {trained_dataset}, num_classes: {num_classes}, model_load_path: {model_load_path}")
if mode == 'val':
for i, (x, y) in enumerate(tqdm(val_dataset)):
x = x.to(device)
y = y.to(device)
with torch.no_grad():
y_pred_prob.append(model(x).cpu().numpy())
y_pred.append(np.argmax(model(x).cpu().numpy(), axis=1))
y_true.append(y.cpu().numpy())
y_true = np.concatenate(y_true)
y_pred = np.concatenate(y_pred)
accuracy, f1 = accuracy_score(y_true, y_pred), f1_score(y_true, y_pred, average='macro')
print(f"Accuracy: {accuracy}, F1: {f1}")
# save the result to a json file
with open(os.path.join(model_path, f"{trained_dataset}_{dataset}_{backbone}_result.json"), "w") as f:
json.dump({"test_dataset" : dataset, "trained_dataset" : trained_dataset, "accuracy" : accuracy, "f1" : f1}, f)
elif mode == 'test' and (dataset == 'aptos' or dataset == 'centerlized' or dataset == 'eyepacs'):
# we don't have the ground truth for test dataset, save the prediction to a csv file
with open(os.path.join(model_path, f"{trained_dataset}_{dataset}_{backbone}_result.csv"), "w") as f:
writer = csv.writer(f)
writer.writerow(["id_code", "diagnosis"])
for i, (x, fake_label, image_name) in enumerate(tqdm(val_dataset)):
x = x.to(device)
with torch.no_grad():
y_pred_prob = model(x).cpu().numpy()
y_pred = np.argmax(y_pred_prob, axis=1)
for j in range(len(y_pred)):
writer.writerow([image_name[j], y_pred[j]])
else:
# for other test dataset, we have the ground truth, save the result to a json file
for i, (x, y, image_name) in enumerate(tqdm(val_dataset)):
x = x.to(device)
y = y.to(device)
with torch.no_grad():
y_pred_prob.append(model(x).cpu().numpy())
y_pred.append(np.argmax(model(x).cpu().numpy(), axis=1))
y_true.append(y.cpu().numpy())
y_true = np.concatenate(y_true)
y_pred = np.concatenate(y_pred)
y_pred_prob = np.concatenate(y_pred_prob)
y_pred_prob = torch.softmax(torch.from_numpy(y_pred_prob), dim=1).numpy()
accuracy, f1= accuracy_score(y_true, y_pred), f1_score(y_true, y_pred, average='macro')
print(f"Accuracy: {accuracy}, F1: {f1}")
# save the result to a json file
with open(os.path.join(model_path, f"{trained_dataset}_{dataset}_{backbone}_test_set_result.json"), "w") as f:
json.dump({"test_dataset" : dataset, "trained_dataset" : trained_dataset, "accuracy" : accuracy, "f1" : f1}, f)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--backbone", type=str, default="resnet50")
parser.add_argument("--model_path", type=str, default="none")
parser.add_argument("--dataset", type=str, default="centerlized")
parser.add_argument("--trained_dataset", type=str, default="centerlized")
parser.add_argument("--mode", type=str, default="val")
parser.add_argument("--device", type=str, default="cuda")
args = parser.parse_args()
main(**vars(args))