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
| import torch
import torch.nn as nn
class ECRAR(nn.Module):
"""
ECRAR: 环境上下文感知雷达动作识别
两个并行分支:
1. 人体运动分支: 4D 点云 → 运动特征
2. 场景上下文分支: 雷达背景 → 场景特征
3. 联合推理: 运动 × 场景 → 动作分类
"""
def __init__(self, num_classes=10, hidden_dim=256):
super().__init__()
self.motion_encoder = MotionEncoder4D(
in_channels=5,
hidden_dim=hidden_dim
)
self.context_encoder = SceneContextEncoder(
in_channels=5,
hidden_dim=hidden_dim
)
self.fusion = nn.Sequential(
nn.Linear(hidden_dim * 2, hidden_dim),
nn.GELU(),
nn.Dropout(0.2),
nn.Linear(hidden_dim, num_classes)
)
def forward(self, human_points, scene_points):
"""
Args:
human_points: (B, N_h, 5) 人体雷达点
scene_points: (B, N_s, 5) 场景雷达点
"""
motion_feat = self.motion_encoder(human_points)
context_feat = self.context_encoder(scene_points)
combined = torch.cat([motion_feat, context_feat], dim=-1)
logits = self.fusion(combined)
return logits
class MotionEncoder4D(nn.Module):
"""4D 雷达人体运动编码器"""
def __init__(self, in_channels=5, hidden_dim=256):
super().__init__()
self.mlp = nn.Sequential(
nn.Linear(in_channels, 64), nn.ReLU(),
nn.Linear(64, 128), nn.ReLU(),
nn.Linear(128, hidden_dim)
)
self.temporal = nn.GRU(
hidden_dim, hidden_dim // 2,
batch_first=True, bidirectional=True
)
def forward(self, points):
feat = self.mlp(points)
feat = feat.max(dim=1)[0]
if feat.dim() == 3:
feat, _ = self.temporal(feat)
feat = feat[:, -1, :]
return feat
class SceneContextEncoder(nn.Module):
"""场景上下文编码器"""
def __init__(self, in_channels=5, hidden_dim=256):
super().__init__()
self.encoder = nn.Sequential(
nn.Linear(in_channels, 64), nn.ReLU(),
nn.Linear(64, 128), nn.ReLU(),
nn.Linear(128, hidden_dim)
)
self.spatial_attn = nn.Sequential(
nn.Linear(hidden_dim, 1),
nn.Softmax(dim=1)
)
def forward(self, scene_points):
feat = self.encoder(scene_points)
attn = self.spatial_attn(feat)
feat = (feat * attn).sum(dim=1)
return feat
if __name__ == "__main__":
model = ECRAR(num_classes=10)
human = torch.randn(2, 100, 5)
scene = torch.randn(2, 500, 5)
out = model(human, scene)
print(f"Human points: {human.shape}")
print(f"Scene points: {scene.shape}")
print(f"Output: {out.shape}")
print(f"Params: {sum(p.numel() for p in model.parameters()):,}")
|
