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
| """
EyeCue论文复现 - 完整训练脚本
论文：Driver Cognitive Distraction Detection via Gaze-Empowered Egocentric Video Understanding
作者：Yoon et al. (2026)
"""
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import Dataset, DataLoader
import numpy as np
from pathlib import Path
class CogDriveDataset(Dataset):
"""CogDrive认知分心数据集"""
def __init__(self, data_root, split='train'):
self.data_root = Path(data_root)
self.split = split
self.annotations = self._load_annotations()
print(f"加载{split}集：{len(self.annotations)}样本")
def _load_annotations(self):
"""加载标注文件"""
annotations = []
for i in range(100 if self.split == 'train' else 20):
annotations.append({
'video_path': f'video_{i:04d}.mp4',
'gaze_path': f'gaze_{i:04d}.npy',
'label': np.random.randint(2),
'scene': np.random.choice(['urban', 'highway', 'rural'])
})
return annotations
def __len__(self):
return len(self.annotations)
def __getitem__(self, idx):
ann = self.annotations[idx]
video = torch.randn(30, 3, 224, 224)
gaze = torch.randn(30, 6)
label = ann['label']
return {
'video': video,
'gaze': gaze,
'label': torch.tensor(label, dtype=torch.long),
'scene': ann['scene']
}
def train_epoch(model, dataloader, optimizer, criterion, device):
"""训练一个epoch"""
model.train()
total_loss = 0
correct = 0
total = 0
for batch in dataloader:
video = batch['video'].to(device)
gaze = batch['gaze'].to(device)
labels = batch['label'].to(device)
optimizer.zero_grad()
logits = model(video, gaze)
loss = criterion(logits, labels)
loss.backward()
optimizer.step()
total_loss += loss.item()
_, predicted = logits.max(1)
total += labels.size(0)
correct += predicted.eq(labels).sum().item()
return total_loss / len(dataloader), correct / total
def validate(model, dataloader, criterion, device):
"""验证"""
model.eval()
total_loss = 0
correct = 0
total = 0
scene_correct = {}
scene_total = {}
with torch.no_grad():
for batch in dataloader:
video = batch['video'].to(device)
gaze = batch['gaze'].to(device)
labels = batch['label'].to(device)
scenes = batch['scene']
logits = model(video, gaze)
loss = criterion(logits, labels)
total_loss += loss.item()
_, predicted = logits.max(1)
total += labels.size(0)
correct += predicted.eq(labels).sum().item()
for i, scene in enumerate(scenes):
if scene not in scene_correct:
scene_correct[scene] = 0
scene_total[scene] = 0
scene_total[scene] += 1
if predicted[i] == labels[i]:
scene_correct[scene] += 1
scene_acc = {k: scene_correct[k]/scene_total[k] for k in scene_total}
return total_loss / len(dataloader), correct / total, scene_acc
if __name__ == "__main__":
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(f"使用设备: {device}")
train_dataset = CogDriveDataset('./data/cogdrive', split='train')
val_dataset = CogDriveDataset('./data/cogdrive', split='val')
train_loader = DataLoader(train_dataset, batch_size=4, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=4, shuffle=False)
model = EyeCue().to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.AdamW(model.parameters(), lr=1e-4, weight_decay=0.01)
num_epochs = 50
best_acc = 0
for epoch in range(num_epochs):
train_loss, train_acc = train_epoch(model, train_loader, optimizer, criterion, device)
val_loss, val_acc, scene_acc = validate(model, val_loader, criterion, device)
print(f"\nEpoch {epoch+1}/{num_epochs}")
print(f"Train Loss: {train_loss:.4f}, Acc: {train_acc:.4f}")
print(f"Val Loss: {val_loss:.4f}, Acc: {val_acc:.4f}")
print(f"Scene Acc: {scene_acc}")
if val_acc > best_acc:
best_acc = val_acc
torch.save(model.state_dict(), 'eyecue_best.pth')
print(f"保存最佳模型，准确率: {best_acc:.4f}")
print(f"\n训练完成！最佳准确率: {best_acc:.4f}")
|