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
| """
疲劳检测数据集评估工具
评估指标:
- 准确率/精确率/召回率/F1
- 帧率(FPS) vs 准确率权衡
- 模型大小 vs 部署可行性
"""
import numpy as np
from typing import Dict, List, Tuple
class FatigueBenchmark:
"""
疲劳检测基准评估
对比多个模型在NTHU-DDD/YawDD/SUST-DDD上的表现
"""
BENCHMARKS = {
'NTHU-DDD': {
'subjects': 36,
'classes': 2,
'fps': 30,
'resolution': (640, 480),
},
'YawDD': {
'subjects': 34,
'classes': 3,
'fps': 30,
'resolution': (640, 480),
},
'SUST-DDD': {
'subjects': 10,
'classes': 2,
'fps': 30,
'resolution': (1280, 720),
'modalities': ['rgb', 'eeg', 'ecg'],
},
}
@staticmethod
def evaluate_model(model_name: str,
accuracy: float,
params_m: float,
fps: int,
dataset: str) -> Dict:
"""
评估模型在各维度的表现
Args:
model_name: 模型名
accuracy: 准确率 0-1
params_m: 参数量(百万)
fps: 推理帧率
dataset: 数据集名
"""
deploy_score = 0
if params_m < 5:
deploy_score += 40 量化友好
elif params_m < 20:
deploy_score += 25
elif params_m < 50:
deploy_score += 10
if fps >= 30:
deploy_score += 40 量产帧率要求
elif fps >= 15:
deploy_score += 25
elif fps >= 10:
deploy_score += 10
if accuracy > 0.85:
deploy_score += 20
elif accuracy > 0.75:
deploy_score += 10
elif accuracy > 0.65:
deploy_score += 5
if deploy_score >= 80:
recommendation = 'production_ready'
elif deploy_score >= 60:
recommendation = 'prototype_ready'
elif deploy_score >= 40:
recommendation = 'research_only'
else:
recommendation = 'not_suitable'
return {
'model': model_name,
'dataset': dataset,
'accuracy': f'{accuracy:.1%}',
'params': f'{params_m:.1f}M',
'fps': fps,
'deploy_score': deploy_score,
'recommendation': recommendation,
'model_size_mb': params_m * 4,
'model_size_int8_mb': params_m * 1,
}
if __name__ == "__main__":
benchmark = FatigueBenchmark()
models = [
('EffRes-DrowsyNet', 0.847, 30, 25),
('Fisher-Gabor-LDCRF', 0.863, 0.1, 30),
('CNN-Baseline', 0.731, 5, 60),
('ViT-Base', 0.825, 85, 15),
('MobileNetV3-LSTM', 0.798, 2, 45),
]
print("=== 疲劳检测模型基准对比 ===\n")
print(f"{'模型':<20} {'准确率':<8} {'参数':<8} {'FPS':<5} "
f"{'部署分':<6} {'推荐':<15} {'INT8大小':<10}")
print("-" * 80)
for name, acc, params, fps in models:
result = benchmark.evaluate_model(name, acc, params, fps, 'NTHU-DDD')
print(f"{name:<20} {result['accuracy']:<8} {result['params']:<8} "
f"{result['fps']:<5} {result['deploy_score']:<6} "
f"{result['recommendation']:<15} "
f"{result['model_size_int8_mb']:.1f}MB")
print(f"\n=== 量产推荐 ===")
for name, acc, params, fps in models:
result = benchmark.evaluate_model(name, acc, params, fps, 'NTHU-DDD')
if result['recommendation'] == 'production_ready':
print(f"✅ {name}: 准确率{acc:.1%}, INT8={params:.1f}MB, {fps}fps")
elif result['recommendation'] == 'prototype_ready':
print(f"⚠️ {name}: 可原型验证，需优化")
|