Joseph Redmon
2015-09-09 393dc8eb6f3a9dd92ec665200444186c1addc5d2
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
177
178
179
180
181
182
183
184
185
186
187
188
#include "region_layer.h"
#include "activations.h"
#include "softmax_layer.h"
#include "blas.h"
#include "box.h"
#include "cuda.h"
#include "utils.h"
#include <stdio.h>
#include <assert.h>
#include <string.h>
#include <stdlib.h>
 
region_layer make_region_layer(int batch, int inputs, int n, int side, int classes, int coords, int rescore)
{
    region_layer l = {0};
    l.type = REGION;
    
    l.n = n;
    l.batch = batch;
    l.inputs = inputs;
    l.classes = classes;
    l.coords = coords;
    l.rescore = rescore;
    l.side = side;
    assert(side*side*l.coords*l.n == inputs);
    l.cost = calloc(1, sizeof(float));
    int outputs = l.n*5*side*side;
    l.outputs = outputs;
    l.output = calloc(batch*outputs, sizeof(float));
    l.delta = calloc(batch*inputs, sizeof(float));
    #ifdef GPU
    l.output_gpu = cuda_make_array(l.output, batch*outputs);
    l.delta_gpu = cuda_make_array(l.delta, batch*inputs);
#endif
 
    fprintf(stderr, "Region Layer\n");
    srand(0);
 
    return l;
}
 
void forward_region_layer(const region_layer l, network_state state)
{
    int locations = l.side*l.side;
    int i,j;
    for(i = 0; i < l.batch*locations; ++i){
        for(j = 0; j < l.n; ++j){
            int in_index =  i*l.n*l.coords + j*l.coords;
            int out_index = i*l.n*5 + j*5;
 
            float prob =  state.input[in_index+0];
            float x =     state.input[in_index+1];
            float y =     state.input[in_index+2];
            float w =     state.input[in_index+3];
            float h =     state.input[in_index+4];
            /*
            float min_w = state.input[in_index+5];
            float max_w = state.input[in_index+6];
            float min_h = state.input[in_index+7];
            float max_h = state.input[in_index+8];
            */
 
            l.output[out_index+0] = prob;
            l.output[out_index+1] = x;
            l.output[out_index+2] = y;
            l.output[out_index+3] = w;
            l.output[out_index+4] = h;
 
        }
    }
    if(state.train){
        float avg_iou = 0;
        int count = 0;
        *(l.cost) = 0;
        int size = l.inputs * l.batch;
        memset(l.delta, 0, size * sizeof(float));
        for (i = 0; i < l.batch*locations; ++i) {
 
            for(j = 0; j < l.n; ++j){
                int in_index = i*l.n*l.coords + j*l.coords;
                l.delta[in_index+0] = .1*(0-state.input[in_index+0]);
            }
 
            int truth_index = i*5;
            int best_index = -1;
            float best_iou = 0;
            float best_rmse = 4;
 
            int bg = !state.truth[truth_index];
            if(bg) continue;
 
            box truth = {state.truth[truth_index+1], state.truth[truth_index+2], state.truth[truth_index+3], state.truth[truth_index+4]};
            truth.x /= l.side;
            truth.y /= l.side;
 
            for(j = 0; j < l.n; ++j){
                int out_index = i*l.n*5 + j*5;
                box out = {l.output[out_index+1], l.output[out_index+2], l.output[out_index+3], l.output[out_index+4]};
 
                //printf("\n%f %f %f %f %f\n", l.output[out_index+0], out.x, out.y, out.w, out.h);
 
                out.x /= l.side;
                out.y /= l.side;
 
                float iou  = box_iou(out, truth);
                float rmse = box_rmse(out, truth);
                if(best_iou > 0 || iou > 0){
                    if(iou > best_iou){
                        best_iou = iou;
                        best_index = j;
                    }
                }else{
                    if(rmse < best_rmse){
                        best_rmse = rmse;
                        best_index = j;
                    }
                }
            }
            printf("%d", best_index);
            //int out_index = i*l.n*5 + best_index*5;
            //box out = {l.output[out_index+1], l.output[out_index+2], l.output[out_index+3], l.output[out_index+4]};
            int in_index =  i*l.n*l.coords + best_index*l.coords;
 
            l.delta[in_index+0] = (1-state.input[in_index+0]);
            l.delta[in_index+1] = state.truth[truth_index+1] - state.input[in_index+1];
            l.delta[in_index+2] = state.truth[truth_index+2] - state.input[in_index+2];
            l.delta[in_index+3] = state.truth[truth_index+3] - state.input[in_index+3];
            l.delta[in_index+4] = state.truth[truth_index+4] - state.input[in_index+4];
            /*
            l.delta[in_index+5] = 0 - state.input[in_index+5];
            l.delta[in_index+6] = 1 - state.input[in_index+6];
            l.delta[in_index+7] = 0 - state.input[in_index+7];
            l.delta[in_index+8] = 1 - state.input[in_index+8];
            */
 
            /*
            float x =     state.input[in_index+1];
            float y =     state.input[in_index+2];
            float w =     state.input[in_index+3];
            float h =     state.input[in_index+4];
            float min_w = state.input[in_index+5];
            float max_w = state.input[in_index+6];
            float min_h = state.input[in_index+7];
            float max_h = state.input[in_index+8];
            */
 
 
            avg_iou += best_iou;
            ++count;
        }
        printf("\nAvg IOU: %f %d\n", avg_iou/count, count);
    }
}
 
void backward_region_layer(const region_layer l, network_state state)
{
    axpy_cpu(l.batch*l.inputs, 1, l.delta, 1, state.delta, 1);
    //copy_cpu(l.batch*l.inputs, l.delta, 1, state.delta, 1);
}
 
#ifdef GPU
 
void forward_region_layer_gpu(const region_layer l, network_state state)
{
    float *in_cpu = calloc(l.batch*l.inputs, sizeof(float));
    float *truth_cpu = 0;
    if(state.truth){
        truth_cpu = calloc(l.batch*l.outputs, sizeof(float));
        cuda_pull_array(state.truth, truth_cpu, l.batch*l.outputs);
    }
    cuda_pull_array(state.input, in_cpu, l.batch*l.inputs);
    network_state cpu_state;
    cpu_state.train = state.train;
    cpu_state.truth = truth_cpu;
    cpu_state.input = in_cpu;
    forward_region_layer(l, cpu_state);
    cuda_push_array(l.output_gpu, l.output, l.batch*l.outputs);
    cuda_push_array(l.delta_gpu, l.delta, l.batch*l.inputs);
    free(cpu_state.input);
    if(cpu_state.truth) free(cpu_state.truth);
}
 
void backward_region_layer_gpu(region_layer l, network_state state)
{
    axpy_ongpu(l.batch*l.inputs, 1, l.delta_gpu, 1, state.delta, 1);
    //copy_ongpu(l.batch*l.inputs, l.delta_gpu, 1, state.delta, 1);
}
#endif