File size: 5,160 Bytes
1cdc47e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import absolute_import, division

import math
import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np


class SemCHGraphConv(nn.Module):
    """
    Semantic channel-wise graph convolution layer
    """

    def __init__(self, in_features, out_features, adj, bias=False):
        super(SemCHGraphConv, self).__init__()
        self.in_features = in_features
        self.out_features = out_features

        self.W = nn.Parameter(torch.zeros(size=(2, in_features, out_features), dtype=torch.float))
        nn.init.xavier_uniform_(self.W.data, gain=1.414)

        self.adj = adj.unsqueeze(0).repeat(out_features, 1, 1)
        self.m = (self.adj > 0)
        self.e = nn.Parameter(torch.zeros(out_features, len(self.m[0].nonzero()), dtype=torch.float))
        nn.init.constant_(self.e.data, 1)

        if bias:
            self.bias = nn.Parameter(torch.zeros(out_features, dtype=torch.float))
            stdv = 1. / math.sqrt(self.W.size(1))
            self.bias.data.uniform_(-stdv, stdv)
        else:
            self.register_parameter('bias', None)

    def forward(self, input):
        # input: (B, T, J, C)
        h0 = torch.matmul(input, self.W[0]).unsqueeze(2).transpose(2, 4)  # B * T * C * J * 1
        h1 = torch.matmul(input, self.W[1]).unsqueeze(2).transpose(2, 4)  # B * T * C * J * 1

        adj = -9e15 * torch.ones_like(self.adj).to(input.device)  # C * J * J
        adj[self.m] = self.e.view(-1)
        adj = F.softmax(adj, dim=2)

        E = torch.eye(adj.size(1), dtype=torch.float).to(input.device)
        E = E.unsqueeze(0).repeat(self.out_features, 1, 1)  # C * J * J

        output = torch.matmul(adj * E, h0) + torch.matmul(adj * (1 - E), h1)
        output = output.transpose(2, 4).squeeze(2)

        if self.bias is not None:
            return output + self.bias.view(1, 1, -1)
        else:
            return output

    def __repr__(self):
        return self.__class__.__name__ + ' (' + str(self.in_features) + ' -> ' + str(self.out_features) + ')'


class LocalGraph(nn.Module):
    def __init__(self, adj, input_dim, output_dim, dropout=None):
        super(LocalGraph, self).__init__()
        
        num_joints = adj.shape[0]

        # Human3.6M
        if num_joints == 17:
            distal_joints = [3, 6, 10, 13, 16]
            joints_left = [4, 5, 6, 11, 12, 13]
            joints_right = [1, 2, 3, 14, 15, 16]

        # Human3.6M detected from Stacked Hourglass
        elif num_joints == 16:
            distal_joints = [3, 6, 9, 12, 15]
            joints_left = [4, 5, 6, 10, 11, 12]
            joints_right = [1, 2, 3, 13, 14, 15]

        # HumanEva
        elif num_joints == 15:
            distal_joints = [4, 7, 10, 13]
            joints_left = [2, 3, 4, 8, 9, 10]
            joints_right = [5, 6, 7, 11, 12, 13]

        # Human3.6M including toe keypoints
        elif num_joints == 19:
            distal_joints = [3, 4, 7, 8, 12, 15, 18]
            joints_left = [5, 6, 7, 8, 13, 14, 15]
            joints_right = [1, 2, 3, 4, 16, 17, 18]

        else:
            raise KeyError("The dimension of adj matrix is wrong!")

        adj_sym = torch.zeros_like(adj)
        for i in range(num_joints):
            for j in range(num_joints):
                if i == j:
                    adj_sym[i][j] = 1
                if i in joints_left:
                    index = joints_left.index(i)
                    adj_sym[i][joints_right[index]] = 1.0
                if i in joints_right:
                    index = joints_right.index(i)
                    adj_sym[i][joints_left[index]] = 1.0

        adj_1st_order = adj.matrix_power(1)
        for i in np.arange(num_joints):
            if i in distal_joints:
                adj_1st_order[i] = 0

        adj_2nd_order = adj.matrix_power(2)
        for i in np.arange(num_joints):
            if i not in distal_joints:
                adj_2nd_order[i] = 0

        adj_con = adj_1st_order + adj_2nd_order

        self.gcn_sym = SemCHGraphConv(input_dim, output_dim, adj_sym)
        self.bn_1 = nn.BatchNorm2d(output_dim, momentum=0.1)
        self.gcn_con = SemCHGraphConv(input_dim, output_dim, adj_con)
        self.bn_2 = nn.BatchNorm2d(output_dim, momentum=0.1)
        self.relu = nn.ReLU()

        self.cat_conv = nn.Conv2d(2*output_dim, output_dim, 1, bias=False)
        self.cat_bn = nn.BatchNorm2d(output_dim, momentum=0.1)

        if dropout is not None:
            self.dropout = nn.Dropout(dropout)
        else:
            self.dropout = None

    def forward(self, input):
        # x: (B, T, K, C)
        x = self.gcn_sym(input)
        y = self.gcn_con(input)

        # x: (B, T, K, C) --> (B, C, T, K)
        x = x.permute(0, 3, 1, 2)
        y = y.permute(0, 3, 1, 2)

        x = self.relu(self.bn_1(x))
        y = self.relu(self.bn_2(y))

        output = torch.cat((x, y), dim=1)
        output = self.cat_bn(self.cat_conv(output))

        if self.dropout is not None:
            output = self.dropout(self.relu(output))
        else:
            output = self.relu(output)
        output = output.permute(0, 2, 3, 1)

        return output